feat(tts): add MiniMax Speech provider

[RayJiang4S]

Summary

• add a MiniMax Speech HTTP T2A provider for OpenMontage TTS workflows
• add MiniMax voice discovery and a batch audition helper for practical voice selection
• document env setup, provider behavior, and MiniMax-specific agent usage

Verification

• python -m pytest tests/tools/test_minimax_tts.py tests/contracts/test_phase3_contracts.py -q -> 75 passed
• registry discovery confirms minimax_tts is registered with capability tts
• TTSSelector(preferred_provider="minimax") is covered by a mocked end-to-end provider routing test
• real MiniMax get_voice call authenticated successfully and returned system voices; paid T2A generation could not be smoke-tested because the account returned insufficient balance

Notes

• This PR covers HTTP T2A generation, metadata capture, voice discovery, and audition helpers.
• Async T2A, WebSocket streaming, voice cloning, and voice design remain future enhancements.