feat: add MiniMax Cloud TTS provider

[octo-patch]

Summary

This PR adds MiniMax Cloud TTS as a first-class text-to-speech provider for joinly, complementing the existing Kokoro, ElevenLabs, and Deepgram options.

What changed

• joinly/services/tts/minimax.py — MinimaxTTS class using the MiniMax t2a_v2 API with PCM output, so audio is consumed directly by the pipeline without additional decoding
• README.md — Added --tts minimax to the Providers section; updated the TTS feature bullet
• tests/services/tts/test_minimax.py — 19 tests: 7 init tests, 9 stream tests (mocked), 3 live integration tests

Usage

# Set your MiniMax API key
export MINIMAX_API_KEY=your-key

# Use MiniMax TTS with default settings (speech-02-hd, English_Graceful_Lady voice)
--tts minimax

# Choose a different model or voice
--tts minimax --tts-arg model=speech-02-turbo --tts-arg voice_id=English_Persuasive_Man

Key design choices

• Requests PCM format (audio_setting.format: pcm) so joinly receives raw 16-bit samples — no MP3 decoding needed
• Default sample rate 32 000 Hz (native MiniMax PCM output rate)
• Voice auto-selected per JOINLY_LANGUAGE (English, Chinese, German, French, Spanish, Japanese, Korean)
• Follows the same aiohttp + AsyncIterator[bytes] pattern as ResembleTTS
• Requires MINIMAX_API_KEY environment variable; raises ValueError on startup if missing

Test plan

• 7 unit tests for __init__ (API key guard, audio_format, voice defaults, model selection)
• 9 stream tests: happy path, chunking, HTTP error, API error, empty audio, network error, payload structure, voice_setting structure, auth header
• 3 live integration tests (skipped when MINIMAX_API_KEY unset, all pass with real key)

tests/services/tts/test_minimax.py - 19 passed