perf(scroll): pace frames to the panel, not to a fixed sleep

Scrolling ran at 44-46 fps on a 2x128x64 chain and 14-17% of frames took
41-53ms, which reads as judder. Four independent causes, each measured on
the hardware; details and the diagnostic recipe are in
docs/SCROLL_PERFORMANCE.md.

The high-FPS loop slept a flat 8ms after every render. display() has
already blocked on the panel's vsync by then, so that sleep was added to a
wait that had happened: ~4ms of render plus 8ms put each iteration at ~12ms
against a 10ms refresh grid, so every swap missed a refresh and the loop
settled at 50fps while asking for 125 -- with no headroom, so a further
14% of frames slipped again. It now sleeps only the remainder, with a 1ms
floor so plugin threads still get the GIL.

ScrollHelper stepped position on a wall clock at 1/scroll_delay steps per
second. Plugins set scroll_delay to the frame period, so that comparison
sat exactly on its own threshold: a frame arriving a hair early moved zero
pixels and rendered an identical frame, dirty-tracking skipped the swap, it
returned in ~2ms, and the beat repeated. No scroll_delay value tunes that
out -- a shorter delay trades stalled frames for periodic double-steps.
Both modes now accumulate elapsed time at the same configured speed, so
position stays proportional to real time.

Sub-pixel blending goes back to off by default. It renders a half-step by
mixing two adjacent columns, which on a coarse panel showing pixel-font
text alternates crisp and smeared frames and reads as shimmer -- visibly
worse than integer stepping on the hardware. Vegas mode still opts in.

disk_cache uses orjson when importable, falling back to the stdlib. Encoding
a ~1MB record drops from 14.8ms to 5.4ms end-to-end, and that work holds the
GIL while a marquee is on screen. display_manager also checksummed the whole
framebuffer twice per frame (dirty tracking, then the preview snapshot); the
snapshot now takes the checksum the caller already computed.

New src/common/scroll_config.py resolves scroll settings in one place. Five
ticker plugins each hand-rolled this and disagreed: odds-ticker ranked the
deprecated scroll_pixels_per_second above the documented scroll_speed/delay
pair, and because that key carries a schema default the documented settings
were dead for every user (ChuckBuilds/ledmatrix-plugins#408), while
ledmatrix-leaderboard read the same key only as a fallback. The resolver also
warns when a speed will not advance a whole number of pixels per refresh,
which is the property that actually determines whether a scroll looks smooth.

scripts/build_rgbmatrix_nogil.sh rebuilds the rgbmatrix binding so it
releases the GIL. Upstream declares SwapOnVSync without nogil, unlike
SetPixel/Clear/Fill beside it, so the render thread held the GIL for the
whole vsync wait and starved background threads into long uninterruptible
bursts. The script patches, builds and self-verifies into a scratch tree;
--install backs up the original and rolls back if the service does not come
back healthy.

Measured after: 100 fps locked, no stalls observed, render thread down from
51% to 19% of one core.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/SCROLL_PERFORMANCE.md

@@ -0,0 +1,188 @@
+# Scroll Performance
+
+How scrolling is paced on this hardware, what was wrong with it, and how to
+configure a plugin so its marquee is smooth.
+
+Measured on a Raspberry Pi 4 driving a 2×128×64 chain (256×64 logical) at
+`limit_refresh_rate_hz: 100`. Numbers below come from that panel.
+
+| | before | after |
+|---|---|---|
+| scroll frame rate | 44–46 fps | **100 fps, locked** |
+| frames ≥ 45 ms | 14–17% | none observed |
+| dominant frame time | 20 ms | **10 ms** |
+| disk cache write (~1 MB) | 14.8 ms | **5.4 ms** |
+
+---
+
+## The one rule that matters
+
+**Motion is smooth when the strip advances a whole number of pixels per panel
+refresh.**
+
+On a 100 Hz panel the crisp speeds are 100 px/s, 200 px/s, 300 px/s. A speed
+that does not divide evenly has to do one of two bad things:
+
+- **blend** two adjacent columns to render a half-step — on pixel-font text
+  this alternates crisp and smeared frames and reads as shimmer, or as the
+  text jumping a pixel ahead of itself;
+- **repeat** a frame — the strip stands still, then jumps, which reads as
+  judder.
+
+Neither is tunable away. Pick a speed that divides evenly.
+
+`src.common.scroll_config.resolve()` warns when a configured speed will not,
+and names the nearest speed that will.
+
+## Configuring a plugin
+
+Use the shared resolver rather than reading config keys yourself:
+
+```python
+from src.common import scroll_config
+
+settings = scroll_config.configure(
+    self.scroll_helper,
+    plugin_config=self.config,
+    global_config=self.global_config,
+    refresh_hz=scroll_config.refresh_hz_from_config(self.global_config),
+    plugin_logger=self.logger,
+)
+```
+
+It resolves every config shape in one place, applies the speed, and returns
+what it did. Precedence, highest first:
+
+1. `display_options.scroll_speed` + `scroll_delay` — **the recommended form**
+2. `display.scroll_speed` + `scroll_delay` — deprecated shape
+3. `scroll_speed` + `scroll_delay` at the root — legacy flat
+4. `scroll_pixels_per_second` — deprecated
+5. the global `display` block
+6. the built-in default (100 px/s)
+
+`scroll_speed` is pixels per frame and `scroll_delay` is the frame period in
+seconds, so the pair means `scroll_speed / scroll_delay` px/s. The recommended
+config for a 100 Hz panel:
+
+```json
+"display_options": { "scroll_speed": 1.0, "scroll_delay": 0.01 }
+```
+
+### Why the deprecated key ranks below the explicit pair
+
+Because some plugins give `scroll_pixels_per_second` a **schema default**, and
+schema defaults are merged into plugin config. Ranking it above the pair means
+it is always present and always wins, so the documented settings become
+unreachable. That is a real, shipped bug — see
+[ledmatrix-plugins#408](https://github.com/ChuckBuilds/ledmatrix-plugins/issues/408).
+
+If you are writing a plugin: do not give a deprecated key a schema default.
+
+## What was actually wrong
+
+Four independent faults, each found by measurement.
+
+### 1. The frame loop slept on top of a wait it had already done
+
+`display_controller.py` ran the high-FPS loop as `render → SwapOnVSync (blocks
+to the panel's refresh) → time.sleep(0.008) → plugin ticks`. The sleep was
+unconditional and added to a wait that had already happened. Render work
+measured ~4 ms, so each iteration cost ~12 ms against a 10 ms refresh grid —
+every swap missed a refresh and landed on the next one. The loop settled at
+exactly 50 fps while asking for 125, with no headroom, so ~14% of frames
+slipped a further refresh.
+
+Now the loop sleeps only the remainder of the frame budget, with a 1 ms floor
+so plugin threads still get the GIL.
+
+### 2. `SwapOnVSync` held the GIL while blocking
+
+The rgbmatrix binding declares it without `nogil` (unlike `SetPixel`, `Clear`
+and `Fill` immediately above it in `cppinc.pxd`), so the render thread held the
+GIL for the entire vsync wait — most of every frame. Background threads were
+starved into long uninterruptible bursts; a 1.5 MB API response costs ~17 ms to
+parse and ~18 ms to re-encode for the cache, and `json.raw_decode` cannot be
+preempted mid-document. Those bursts are what the render loop then waited on.
+
+Fixed by rebuilding the binding: `scripts/build_rgbmatrix_nogil.sh`.
+
+### 3. Sub-pixel blending was wrong for this display
+
+Enabling it made things worse, not better — see the rule at the top. It is off
+by default and only Vegas mode opts in via `set_sub_pixel_scrolling(True)`.
+
+### 4. Frame-based stepping raced the vsync clock
+
+Frame-based mode gated motion on a wall clock at `1/scroll_delay` steps per
+second. Plugins set `scroll_delay` to the frame period, which puts that
+comparison exactly on its own threshold: a frame arriving a hair early moved
+zero pixels and rendered an identical frame, which dirty-tracking skipped, so
+it returned in ~2 ms and the beat repeated. No `scroll_delay` value tunes this
+out — a shorter delay just trades stalled frames for periodic double-steps.
+
+`ScrollHelper` now accumulates elapsed time in both modes at the same
+configured speed, so position stays proportional to real time.
+
+## Diagnosing a juddery scroller
+
+**`Avg FPS` will lie to you.** It is a 100-frame moving average, and a 2 ms
+duplicate frame plus a 21 ms double-wait average to exactly 10 ms. A ticker
+that is stalling on half its frames still reports a healthy `100.0`.
+
+Look at the **distribution** instead:
+
+```bash
+journalctl -u ledmatrix --since "-10min" --no-pager \
+  | grep -oE "Frame time: [0-9.]+ms" | awk '{print $3}' | sed 's/ms//' \
+  | awk '{printf "%.0f\n", $1}' | sort -n | uniq -c
+```
+
+Reading it, on a 100 Hz panel:
+
+| you see | it means |
+|---|---|
+| everything at 10 ms | healthy |
+| a mode at ~2 ms | **duplicate frames** — the swap was skipped because the image did not change. The scroller is advancing less than one pixel per frame. |
+| a mode at 20/30/50 ms | frames missing refreshes — per-frame work is overrunning, or a background thread is holding the GIL |
+| `Avg FPS` above 100 | duplicates present, unless the scroll cycle has completed and is idling |
+
+Then confirm what the plugin actually loaded — config edits do not always reach
+the running code:
+
+```bash
+journalctl -u ledmatrix --since "-5min" --no-pager | grep -iE "px/s|px/frame"
+```
+
+If a plugin logs its scroll config **twice** with different modes, the second
+line is what is running.
+
+## Rebuilding the binding
+
+```bash
+bash scripts/build_rgbmatrix_nogil.sh              # build into a scratch dir
+sudo bash scripts/build_rgbmatrix_nogil.sh --install
+sudo bash scripts/build_rgbmatrix_nogil.sh --rollback
+```
+
+The build never touches the installed module. `--install` backs up the original
+to `~/rgbmatrix-core.so.ORIGINAL` first, and rolls back automatically if the
+service does not come back healthy. Requires `build-essential`; Cython is
+installed into a cached venv under `~/.cache/ledmatrix-cython`.
+
+Re-run it after upgrading `rpi-rgb-led-matrix`, since a library upgrade
+replaces the patched binding.
+
+## Faster JSON
+
+`src/cache/disk_cache.py` uses `orjson` when it is importable and falls back to
+the stdlib otherwise, so it is optional:
+
+```bash
+sudo pip3 install --break-system-packages orjson
+```
+
+Encoding is where it pays — about 7× on this hardware. Decoding gains far less
+(~1.3× on large payloads) because the cost there is building Python objects,
+not scanning text. That is also why moving parsing to a subprocess does not
+help: `pickle.loads` of the same payload costs 8.1 ms against `json.loads` at
+10.9 ms, so the work just moves rather than disappearing.