Add Rev AI transcription provider

[attaboy11]

Summary

• Add a revai provider for the current provider-registry architecture.
• Submit local audio to Rev AI's async Speech-to-Text API, poll job status, and fetch the completed transcript as plain text.
• Add mocked coverage for upload, polling, result retrieval, availability, missing IDs, failed jobs, and registry discovery.
• Document the Rev AI environment variables and --provider revai CLI usage in the README.

Validation

• .venv/bin/pytest tests/providers/test_revai.py tests/test_registry.py -q
• .venv/bin/pytest -q
• .venv/bin/black --check --target-version py313 sapat/providers/revai.py tests/providers/test_revai.py tests/test_registry.py
• git diff --check
• .venv/bin/python -m compileall -q sapat tests

Supports the Daytona content bounty at daytona/content#13 as the companion Sapat provider implementation.

AI-assisted with OpenAI Codex; I reviewed the diff and local validation before submitting.

[attaboy11]

Follow-up hardening pushed in 498da2f.

What changed:

• Rev AI multipart uploads now send job options through the documented options form field instead of top-level form fields.
• --model fusion, --model low_cost, --model machine, or --model human now maps to Rev AI's transcriber option while the default path stays unchanged.
• Added regression coverage for option construction, automatic language omission, and media content type.
• Updated README usage to reflect the actual CLI flag names and Rev AI model modes.

Validation:

• .venv/bin/python -m pytest -> 185 passed
• .venv/bin/python -m black --check sapat/providers/revai.py tests/providers/test_revai.py tests/test_registry.py -> passed
• git diff --check -> passed

The current Continuous AI statuses all report Agent encountered an error from the external service, so I could not extract actionable check feedback from them.