[Plan] support more tts models

[zhzLuke96]

PLAN desc

支持更多 tts models

由于目前主要基于 chattts 开发，部分模块对其有依赖，需要逐个解离

• 需要区分不同模型的能力问题，涉及 Special tokens and refiners
• 加载多模型可能需要引入自动unload检测
• tn管道需要支持函数可拔插

TODOs

• refactor Speaker 模块: 支持多种模型，并支持内嵌 audio
  • 实现 speaker 序列化反序列化，消除 .pt 模块依赖
  • [Plan] Speaker Card file (PNG embed) #104
• refactor generate 模块 [Plan] Enhanced post-processing pipeline #91
• refactor model 模块 [Plan] Enhanced post-processing pipeline #91
• refactor TN 模块 [Plan] Enhanced post-processing pipeline #91
• add support: fish speech
  • ref: https://github.com/fishaudio/fish-speech
• add support: CosyVoice
  • ref: https://github.com/FunAudioLLM/CosyVoice

[zhzLuke96]

refs #89

[zhzLuke96]

added 1835565 af9efdf c8905df

fish speech 推理部分已接入，但是效果有点感人...😓 暂时不推进了，有兴趣调试的可以在 modules.core.models.tts.FishSpeechModel 模块测试

[zhzLuke96]

done