Fix MaskGCT/Metis dependencies

[jneuendorf-i4h]

✨ Description

⚠️ Remark

I tried to setup my environment for the Metis TTS model with uv. I am aware that you suggest using conda instead.
Even if you discard this PR, I figured it could serve as reference for other users.

Running uv run python -m models.tts.metis.metis_infer_tts after uv add --python 3.10 --requirements models/tts/maskgct/requirements.txt resulted in the following runtime/dependency errors:

1. ImportError: cannot import name 'setLangfilters' from 'LangSegment.LangSegment'
   This is due to an error in the LangSegment library: An import contains removed functions. Previous versions aren't available on PyPI, so I had to patch it manually. See ishine/LangSegment@2ccc10b#r161867267
2. RuntimeError: espeak not installed on your system
   On macOS, espeak-ng installed with Homebrew was not detected for some reason. A solution is specifying PHONEMIZER_ESPEAK_LIBRARY="$(brew --prefix espeak-ng)/lib/libespeak-ng.dylib"
3. ModuleNotFoundError: No module named 'peft' $\leadsto$ missing in requirements
   1. ImportError: cannot import name 'clear_device_cache' from 'accelerate.utils.memory' $\leadsto$ peft==0.14.*
   2. ImportError: cannot import name 'EncoderDecoderCache' from 'transformers' $\leadsto$ peft==0.13.2
4. ModuleNotFoundError: No module named 'langid' $\leadsto$ missing in requirements

This PR targets 3. and 4.

👨‍💻 Changes Proposed

• update models/tts/maskgct/requirements.txt as suggested

✅ Checklist

• Code has been reviewed
• Code complies with the project's code standards and best practices
• Code has passed all tests
• Code does not affect the normal use of existing features
• Code has been commented properly
• Documentation has been updated (if applicable)
• Demo/checkpoint has been attached (if applicable)