feat: add MiniMax Cloud TTS engine (speech-2.8-hd / speech-2.8-turbo)

[octo-patch]

Summary

This PR adds MiniMax Cloud TTS as a first-class TTS engine in VoxNovel, alongside the existing local Coqui / StyleTTS2 models.

Changes

• minimax_tts.py (new): Lightweight client for the MiniMax T2A v2 API. Accepts text + voice ID, decodes hex-encoded MP3 response, converts to WAV via ffmpeg. Supports speech-2.8-hd and speech-2.8-turbo models.
• 2GPU_Audio_generation.py: "MiniMax TTS" added to TTS model combobox; 12 verified English voice IDs added to character voice-actor dropdowns; generate_audio() dispatches to MiniMax API when selected.
• headless_voxnovel.py: Same three changes for the headless / Colab path.
• tests/test_minimax_tts.py (new): 24 unit tests (HTTP + subprocess mocked) + 3 integration tests; all 27 pass.

How to use

1. Set your API key: export MINIMAX_API_KEY=your_key_here
2. Launch VoxNovel and select MiniMax TTS in the TTS model dropdown.
3. Each character voice-actor dropdown now includes minimax:English_Graceful_Lady, minimax:Deep_Voice_Man, etc.
4. Assign a MiniMax voice to each character and click Generate.

Available voices

English_Graceful_Lady, English_Insightful_Speaker, English_radiant_girl, English_Persuasive_Man, English_Lucky_Robot, Wise_Woman, cute_boy, lovely_girl, Friendly_Person, Inspirational_girl, Deep_Voice_Man, sweet_girl

Why MiniMax TTS?

• No GPU required - great for machines without CUDA
• High quality speech-2.8-hd model
• Fast speech-2.8-turbo variant for lower latency
• Zero local model download - ideal for Colab

Test plan

• 24 unit tests pass without network (python -m pytest tests/test_minimax_tts.py -k "not Integration")
• 3 live integration tests pass with MINIMAX_API_KEY set
• All 27 tests pass