feat(meta): add Muse Voice streaming STT plugin

[seyeong-han]

Summary

Adds livekit-plugins-meta with streaming speech-to-text support for Meta Muse Voice Transcribe 1.0.

• exposes livekit.plugins.meta.STT
• sends 24 kHz mono PCM16 audio in paced 80 ms frames
• supports static keywords and language bias at stream creation
• maps cumulative Muse events into non-overlapping LiveKit speech turns
• emits incremental recognition usage from successfully sent audio
• avoids retrying after audio has been consumed, preventing silent speech loss
• redacts credentials, provider payloads, transcripts, and WebSocket close reasons from errors
• adds the workspace source and livekit-agents[meta] extra

The adapter uses LiveKit's shared aiohttp context and the framework-managed input channel/resampler. Batch recognition, diarization, detected-language metadata, and active-stream keyterm updates are intentionally not advertised.

Testing

• uv run pytest tests/test_meta_stt.py --unit (33 passed)
• make check (format, Ruff, and mypy across 648 source files)
• wheel and sdist build, followed by an isolated Python 3.10 wheel import
• real Muse realtime smoke: handshake, paced 24 kHz audio, usage events, endStream, and clean close
• full hermetic run: 2462 passed, 5 skipped; the remaining 9 tests/test_room.py cases independently reproduce a local LiveKit server 401 invalid API key: devkey and are unrelated to this plugin

[CLAassistant]

All committers have signed the CLA.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.