feat(ipc)!: remove the cache-key mailboxes (control socket stage 5)

The control socket is now the only way the web interface sends the display a
command. The display stops reading display_on_demand_request and
plugin_error_clear_request, and the web interface stops writing them.

- Display: no mailbox poll (MailboxWatch, the 1 s / 0.25 s cadence,
  _consume_on_demand_request, the deprecation log) and no persisted
  display_on_demand_processed_id guard; the error publisher reads no clear
  request. CacheManager.file_signature and MailboxWatch are removed.
- A write to either retired key is dropped by CacheManager.save_cache and
  logged once per writer, naming the plugin from the call stack (or the
  request's plugin_id), with the API to move to.
- Web: on-demand start with no display listening starts the service (when
  start_service) and sends the request again once the socket answers (45 s,
  10 s for a running service without a socket yet); every other failure is
  a 503 (400 for invalid_args). Stop answers 503 when no display listens,
  unless stop_service. errors/clear answers 503 with a reason-specific
  message instead of writing a request; clear_pending is always false.
  src.ipc.client.should_fall_back is replaced by display_not_listening.
- Kept: display_current_state, display_on_demand_state,
  plugin_runtime_snapshot and the heartbeat (read whenever the socket cannot
  answer), and display_on_demand_config (the display's resume record).

Tests: mailbox-only tests removed (test_on_demand_mailbox.py, the mailbox
cadence, file_signature and MailboxWatch tests); tests that injected
requests through the mailbox now use the socket queue or a plugin's
in-process request. The run-loop harness sends on-demand requests over its
fake control socket, so four golden traces change: on-demand starts and
stops land at the request instant instead of the next 0.25 s mailbox look
(one frame fewer on the screen they end), and in vegas.json within one
frame instead of 263 ms, which shifts the later 1 s-throttled WiFi-notice
check by under a second.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test/test_run_loop_socket_wake.py

@@ -12,8 +12,8 @@ frame, so a command is applied:
 * at the next frame in Vegas (8 ms here, at 125 fps).
 
 Each test runs the real run() loop on the fake clock of
-test/_run_loop_harness.py and compares the socket with the mailbox for the
-same request. The latencies are the fake clock's, so they are exact.
+test/_run_loop_harness.py. The latencies are the fake clock's, so they are
+exact. (The file mailbox these were once compared with is gone: stage 5.)
 """
 
 import os
@@ -32,13 +32,10 @@ def _event_time(trace, kind):
     return next(e[0] for e in trace["events"] if e[1] == kind)
 
 
-def _on_demand_latency(tmp_path, build, via, horizon=40):
+def _on_demand_latency(tmp_path, build, horizon=40):
     h = RunLoopHarness(tmp_path, horizon=horizon)
     build(h)
-    if via == "socket":
-        h.control_socket().post(POSTED, Command.ON_DEMAND_START, {"plugin_id": "weather"})
-    else:
-        h.on_demand_request(POSTED, "mb1", plugin_id="weather")
+    h.control_socket().post(POSTED, Command.ON_DEMAND_START, {"plugin_id": "weather"})
     trace = h.run()
     return round(_event_time(trace, "on-demand-start") - POSTED, 3), trace
 
@@ -60,18 +57,12 @@ def _vegas(h):
     h.enable_vegas(cycle=30)
 
 
-@pytest.mark.parametrize("build, mailbox_latency", [
-    (_static, 0.7),    # the next 1 s frame, at t=11
-    (_dwell, 0.2),     # the next 0.25 s tick, at t=10.5
-    (_vegas, 0.02),    # the next 10-frame check: 80 ms at 125 fps, ~0.4 s on a Pi 4
-], ids=["static-screen", "dwell", "vegas"])
-def test_a_socket_command_lands_at_once(tmp_path, build, mailbox_latency):
-    (tmp_path / "s").mkdir()
-    (tmp_path / "m").mkdir()
-    socket_latency, trace = _on_demand_latency(tmp_path / "s", build, "socket")
-    via_mailbox, _ = _on_demand_latency(tmp_path / "m", build, "mailbox")
-
-    assert via_mailbox == pytest.approx(mailbox_latency, abs=0.002)
+@pytest.mark.parametrize("build", [_static, _dwell, _vegas],
+                         ids=["static-screen", "dwell", "vegas"])
+def test_a_socket_command_lands_at_once(tmp_path, build):
+    # Before stage 2 these waited for the next 1 s frame, the next 0.25 s
+    # dwell tick, or the next 10-frame Vegas check.
+    socket_latency, trace = _on_demand_latency(tmp_path, build)
     if build is _vegas:
         # One 8 ms frame: Vegas checks the queue every frame now.
         assert socket_latency <= 0.008