Fix SharedFrame mutex panic — handle poisoned locks gracefully

[CREVIOS]

Problem

`SharedFrame` uses `Mutex::lock().unwrap()` in multiple places. If another thread panics while holding the lock, subsequent `.unwrap()` calls will panic too, crashing the entire app.

Affected call sites:

• `begin_write()` — line ~91 in `shared_frame.rs`
• `publish()` — line ~127 in `shared_frame.rs`
• `update_full()` — line ~107 in `shared_frame.rs`
• `wait_for_frame()` — line ~167 in `shared_frame.rs`
• `dimensions()` — line ~173 in `shared_frame.rs`

Fix

Replace all `.lock().unwrap()` with `.lock().unwrap_or_else(|e| e.into_inner())`:

```rust
// Before:
let mut inner = self.inner.lock().unwrap();

// After:
let mut inner = self.inner.lock().unwrap_or_else(|e| {
log::warn!("SharedFrame mutex was poisoned, recovering");
e.into_inner()
});
```

This recovers from a poisoned mutex by extracting the data regardless. The inner operations are infallible memcpy, so the data is still valid even if the previous holder panicked.

Files to modify

• `src-tauri/src/renderer/shared_frame.rs` — all `.lock().unwrap()` calls

Priority: P1 — prevents crash in production