Add changes for streaming mode in TTS models

[yadavsahil197]

Have you read the Contributing Guidelines?

Issue #

Describe your changes

Adding support for new TTS models - orpheus and kokoro along with the streaming mode.

---

Note

Adds binary audio streaming handling in requestor (sync/async) and enhances TTS stream_to_file to assemble and save WAV/MP3/RAW outputs.

• Request handling (src/together/abstract/api_requestor.py):
  • Handle streaming by content type: SSE (text/event-stream) vs binary audio (audio/wav, audio/mpeg, application/octet-stream).
  • Return chunked binary generators for audio streams (sync/async) and preserve bytes for non-streaming binary responses.
  • Refactor async response reading to avoid double reads and properly decode text vs binary.
• Audio TTS (src/together/types/audio_speech.py):
  • AudioSpeechStreamResponse.stream_to_file(file_path, response_format) now:
    • Infers format from filename or parameter; supports wav, mp3, raw.
    • Aggregates streaming chunks (raw bytes or SSE base64) and writes output.
    • Adds WAV header for raw PCM; validates MP3 frames; writes RAW as-is.
  • New helper _write_wav_header to construct valid WAV files for PCM data.

^{Written by Cursor Bugbot for commit 2f6b638. This will update automatically on new commits. Configure here.}