feat: multi-participant rooms with priority routing

[SunnyYadav16]

Replace strict turn-taking with concurrent speech capture and priority-based TTS output routing.

• Expand rooms to support up to 10 simultaneous participants
• Add PriorityRouter: creator (role A) gets high-priority TTS that interrupts all recipients' playback; others queue normally
• Remove TurnStateMachine gating — all speech events always accepted
• Add asyncio.Lock around ASR for CTranslate2 thread safety
• Replace server-side echo suppression (lock_user/mic_locked) with 300ms client-side cooldown after TTS playback ends
• Broadcast translations to all participants via get_others() with per-language TTS caching
• Add tts_interrupt message + priority-aware frontend TTS queue
• Add "Court officer is speaking…" animated banner on high-priority TTS
• Add speaker_role field to transcript messages
• Update README with Phases 6–11 documentation, REST API reference, session recording details, and priority routing architecture