Add Sarvam AI transcription provider

[SadmanPinon]

Summary

• add a Sarvam AI speech-to-text provider using the REST /speech-to-text endpoint
• wire --api sarvam through the CLI and document the required .env settings
• add mocked tests for request construction, missing API key handling, response parsing, and CLI routing

Validation

• /tmp/sapat-sarvam-venv/bin/python -m unittest discover -s tests
• /tmp/sapat-sarvam-venv/bin/python -m py_compile src/sapat/script.py src/sapat/transcription/*.py tests/test_sarvam.py
• /tmp/sapat-sarvam-venv/bin/sapat --help

Context: companion implementation for daytonaio/content#13.\n## Demo\n- Companion walkthrough video: https://github.com/SadmanPinon/content/blob/sarvam-sapat-transcription-guide/guides/assets/20260521_run_sarvam_transcription_with_sapat_demo.mp4\n- Bounty guide PR with the in-repo demo asset: https://github.com/daytonaio/content/pull/235\n

[SadmanPinon]

Note on the current red checks: gh pr checks reports all three Continuous AI statuses as Agent encountered an error, and the Anti-Slop target URL currently returns Cannot GET /api/agent-check-summary/..., so I do not see actionable code feedback from those checks yet.

I re-ran the source validation locally after the latest push:

• PYTHONPATH=src /tmp/sapat-sarvam-venv/bin/python -m unittest discover -s tests — 5 tests passed
• PYTHONPATH=src /tmp/sapat-sarvam-venv/bin/python -m py_compile src/sapat/script.py src/sapat/transcription/*.py tests/test_sarvam.py
• git diff --check

I can address any maintainer or bot feedback if a concrete finding comes through.