Volume HID, bounded simctl, and interaction-paced simulator capture

[jkelleyrtp]

Summary

Three iOS Simulator changes, applied from a patch on top of the ADB socket work:

1. Volume/mute buttons. SimulatorHID gains press_hid_button(page, usage, hold_ms), backed by IndigoHIDMessageForHIDArbitrary, which is dlsym'd optionally — volume up/down/mute are consumer-page usages (0x0c / 0xe9, 0xea, 0xe2) rather than Indigo button IDs. Because the symbol may be absent, the worker reports what it resolved:

   pub fn spawn_input_worker(udid: &str) -> Result<(Sender<InputCommand>, InputCapabilities)>
   //                                              ^ arbitrary_hid: bool

   SimSession::input_capabilities() exposes it so a client can hide the buttons instead of firing commands that fail on the HID thread.

2. xcrun simctl can no longer hang the caller. settings::simctl now spawns and reaps with a 2s deadline (kill + wait on timeout) instead of Command::output(), which had no bound.

3. Capture stays at the configured fps while the user is interacting. The framebuffer idle thread previously always slept IDLE_INTERVAL (200ms), so a screen whose only change came from an in-flight gesture or its animation was re-emitted at ~5fps. SimSession::send_input now calls capture.note_interaction(), which pushes active_until = now + 1s and wakes the thread through a Condvar; while active the tick interval is 1s / config.fps. EncodedFrame/EncodedChunk also carry captured_at from the framebuffer callback through the encoder, so latency can be measured against capture time rather than emit time.

Massaging on top of the patch

• The patch kept a spawn_input_worker wrapper that discarded the capabilities and had no callers; folded it into the single tuple-returning spawn_input_worker.
• The idle loop computed the active/idle interval twice with duplicated bodies; that is now CaptureState::tick_interval(active_until), called once before the wait_timeout and once after it (an interaction can arrive while the thread is parked).
• Restored the first line of the REWIRE_INTERVAL doc comment, which the patch's hunk had eaten.

Testing

cargo fmt --all, cargo clippy --workspace --all-targets -- -D warnings, cargo test --workspace --lib --bins, cargo test -p accessibility-cli --test cli_smoke — all green on Linux.

Most of the diff is macOS-only and does not compile on this box; the x86_64-apple-darwin / aarch64-apple-darwin build jobs are the first real compile of it, and nothing here has run against a booted simulator.

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/d88fc07ea8064b38a4b4e5a7481ffebc
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/d88fc07ea8064b38a4b4e5a7481ffebc?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