Merge dev into main (Update 0.0.1-alpha) - #1

Completes the item-StreamController#1 pattern (issue StreamController#55) in the sibling method: the loop
re-read controller.active_page per check, so a non-None guard could race a
concurrent (dis)connect clearing it and then deref None.json_path. Snapshot
it once, mirroring update_input/update_key_image (item StreamController#8). _pages_lock is
the wrong tool here -- it guards self.pages, not active_page writes.

Also document that _save_locks grows one entry per page path and is never
pruned (bounded by page count; no eviction machinery warranted).

Extends scenario_reload_similar_pages with a flip-to-None probe
(red-checked: pre-fix raises AttributeError on .json_path).

…longer freezes the deck (StreamController#1)

The writer thread (sole path for paints/brightness/Clear/ClearAndClose
since the single-writer migration) ran unguarded: any uncaught exception
killed it, freezing the deck until replug while control_q grew unbounded.

run() now delegates to _run_one_tick() inside a guarded loop: a tick
exception is logged with full traceback (rate-limited to 1/5s with a
suppressed-repeat count -- local limiter until StreamController#91), _stop is honored in
the except path so a failing body can't strand close(), and a 0.25s
wake-event backoff bounds a persistent failure at ~4 retries/s instead of
a 100% spin. @log.catch on run() would log once and let the thread die --
the guard must live inside the while. Composes with StreamController#80: the hooks report
an escaping death; this prevents the death.

The five audit-verified crash vectors are also fixed at the source:
(a) init_inputs is build-then-swap and the screensaver drops its
`inputs = {}` pre-clear (was an empty-dict window on every screensaver
entry); the loop snapshots the dict once per tick and uses .get();
(b) the loop snapshots background.video once (set_video(None) race);
(c) touchscreen get_current_image guards a nulled active_page (blank
strip); (d) the touchscreen tick gate snapshots state.background_video
(_release_background_video race); (e) _exec_clear snapshots the
touchscreen slot and nulls it identity-compared -- the crash half of StreamController#1
and this site's half of StreamController#8.

tests/scenario_writer_survival.py (first leg of StreamController#61) runs the REAL loop
over the stub tier: survives a poisoned tick, resumes paints, caps the
log storm, and stop() joins while every tick raises. fixtures gains
_QuietInputState (get_active_state) since the live loop path never
executed on stubs before (StreamController#69 stub-drift). Suite: 34/34.

…ks, finally'd running flag (StreamController#1 review round 1)

Review round 1 on MR !3 confirmed two starvation holes in the new guard
plus cheap hardening items; all applied:

- Batch recovery [MEDIUM]: a caught tick exception had already popped the
  tick's image_tasks/touchscreen_task, so the failing frame's siblings were
  lost with it and nothing scheduled recovery -- after e.g. a TypeError on
  key 3 of a 15-key repaint, the remaining keys kept the previous page's
  imagery silently forever. The except path now arms
  _schedule_full_repaint(); its existing 2s rate limit keeps a
  deterministic per-tick failure from repaint-storming.
- Drain-first ticks [MEDIUM]: check_resume_gap/_run_pending_repaint ran
  before drain_control_queue inside the same guard, so a persistent
  pre-drain failure starved SetBrightnessMsg/ClearMsg/ClearAndCloseMsg
  (deck never blanked/closed on quit) -- and the 'drains first,
  unconditionally' comment was false. The drain now truly leads the tick;
  comment updated to cover both orderings.
- running=False moved into a finally [LOW]: the guard is except Exception,
  so an escaping BaseException (SystemExit/KeyboardInterrupt) left
  running=True and every later stop() burned its full 2s join timeout.
- Bounded except-path backoff [LOW]: every producer sets _wake_event, so a
  single wait(0.25) under a set_media storm returned immediately and the
  retry rate tracked the producer rate; re-wait until the backoff truly
  elapsed, only _stop cuts it short.
- Reworded the rate-limiter suffix ('N earlier repeats were suppressed') --
  the old 'in the last 5s' wording was wrong when the suffix emitted hours
  after the burst.
- Updated the stale 'loop has no exception guard' comment in
  DeckController.__init__.

Tests: scenario_writer_survival gains batch-recovery and control-drain
legs mirroring the review's probes; both red-tested against their reverted
fixes. Harness 34/34.

fix(deck): guard the sole-writer media loop (StreamController#1)

Closes StreamController#1

See merge request naz/StreamController!3