Sarvam's WebSocket API decoder format mismatch

[DharanHaarish]

Summary

Fix mime_type mismatch in Sarvam TTS WebSocket streaming path. The SynthesizeStream class was declaring audio/wav but Sarvam's WebSocket API actually returns MP3 audio, causing the LiveKit WAV decoder to fail on MPEG bytes.

Changes

• Changed mime_type from "audio/wav" to "audio/mp3" in SynthesizeStream._run()

Notes

• The HTTP REST path (ChunkedStream) correctly handles WAV format
• The WebSocket path was incorrectly assuming the same format without verification
• This aligns with how other plugins (AWS, Resemble) handle MP3 streams

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

DharanHaarish seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[devin-ai-integration]

✅ Devin Review: No Issues Found

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

View in Devin Review to see 1 additional finding.