Add MiniMax text-to-audio provider

[octo-patch]

Reason: Add MiniMax T2A support to the executable speech pipeline.

Changes:

• Add the Global and China HTTP endpoints, the supported speech model catalog, and audio format validation.
• Forward the supported request fields and parse both hex-encoded and URL audio responses.
• Route the MiniMax family through the existing speech processor and document its environment variables.
• Cover endpoint selection, payload construction, response parsing, errors, and processor dispatch with mocked tests.

Checks:

• .venv/bin/pytest -q tests/test_minimax_tts.py (5 passed)
• uvx black --check src/audio/minimax_tts.py tests/test_minimax_tts.py
• uvx ruff check src/audio/minimax_tts.py tests/test_minimax_tts.py
• .venv/bin/python -m compileall -q src/audio/minimax_tts.py src/audio/tts.py tests/test_minimax_tts.py