[fast_inference_] 重构很多api_v2.py，make api great and greater again !

[X-T-E-R]

修改：

1. 改用FileResponse(tmp_file_path, media_type=f"audio/{format}", filename=f"audio.{format}")，替代原本非流式的返回，支持更多格式，速度能更快，并且会避免占线
2. 重写接口函数，避免无用代码

@APP.get("/tts")
@APP.post("/tts")
async def tts_get_endpoint(request: Request):
    # 尝试从JSON中获取数据，如果不是JSON，则从查询参数中获取
    if request.method == "GET":
        data = request.query_params
    else:
        data = await request.json()

3. 利用上了 TTS_Request()类，并使用了更温和的更新规则

class TTS_Request(BaseModel):
    text: str = None
    ……

    def check(self):
        if (self.text_lang in [None, ""]) or self.text_lang.lower() not in tts_config.languages:
            self.text_lang = "auto"
        if (self.prompt_lang in [None, ""]) or self.prompt_lang.lower() not in tts_config.languages:
            self.prompt_lang = "auto"
            
        if self.text in [None, ""]:
            return JSONResponse(status_code=400, content={"message": "text is required"})
        if self.ref_audio_path in [None, ""]:
            # 这里没加验证存在性，TODO
            return JSONResponse(status_code=400, content={"message": "ref_audio_path is required"})
        if self.streaming_mode and self.media_type not in ["wav", "raw", "ogg", "aac"]:
            return JSONResponse(status_code=400, content={"message": f"media_type {self.media_type} is not supported in streaming mode"})
        if self.text_split_method not in cut_method_names:
            return JSONResponse(status_code=400, content={"message": f"text_split_method:{self.text_split_method} is not supported"})
        return None

4. 重写了更新函数，节省时间，节省生命。没给值就继承默认值

def update(self, req:dict):
        for key in req:
            if hasattr(self, key):
                type_ = type(getattr(self, key))
                value = unquote(str(req[key]))
                if type_ == bool:
                    value = value.lower() in ["true", "1"]
                elif type_ == int:
                    value = int(value)
                elif type_ == float:
                    value = float(value)
                setattr(self, key, value)
                

5. 将两个language默认值设置为auto， 因此仅需要2个必要参数：

"text": "",                   # str.(required) text to be synthesized
"ref_audio_path": "",         # str.(required) reference audio path.

[X-T-E-R]

其中部分代码来自GSVI：https://github.com/X-T-E-R/GPT-SoVITS-Inference 已标注

可能存在一些bug，请试一试。