Drive simulator fill captures from a strict GCD timer on a worker queue

[jkelleyrtp]

Summary

Puts the simulator fill cadence on a strict libdispatch timer instead of a sleeping thread, because relative waits (nanosleep / pthread_cond_timedwait) get coalesced to ~100 ms on virtualised hosts, capping the thread-driven fill at ~12 fps.

• New serial worker queue (QoS user-interactive) does the actual capture work. SimulatorKit callbacks now call trigger(force) — a cheap coalescing enqueue (pending/pending_force atomics) — instead of running capture inline on the callback queue, which the render server blocks on.
• start_fill_timer creates a DISPATCH_TIMER_STRICT source on the worker queue via raw dispatch_source_* FFI (dispatch2 0.3 has no safe timer-source wrapper). Handler fires capture(true) only when last_capture.elapsed() + FILL_SLACK >= tick_interval, with 1.5 ms slack so timer jitter doesn't beat a 60 Hz fill down to 30 Hz.
• The old idle thread no longer drives re-emits; it only re-wires the pipeline until the first frame lands.
• stop() cancels/releases the timer and drains the worker with exec_sync before releasing registrations.

macOS-only code path (packages/accessibility-ios-sys/src/macos/framebuffer.rs); verified cargo fmt --check and cargo check --workspace on Linux, macOS compile relies on CI.

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/e327e80d7dee497288adbc3cdc479ac7
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/e327e80d7dee497288adbc3cdc479ac7?variant=devin-insiders
Requested by: @jkelleyrtp

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring