Captions with live transcription and speaker attribution

Meet now has opt-in live captions using Nemotron STT with per-participant attribution, bounded VAD, and subscriber-only delivery.