* fix asr infer.py

* add readme.

This reverts commit 6e0044be.

*merge develop into speechx

* add all whisper model size support

* add choices in parser.

* fix Whisper cli model download path error.

* fix resource download path.

* fix code style

* add all whisper model size support

* add choices in parser.

* add whisper asr large model decoding, test=asr

* fix code style.

* fix json code style.

* remove resource and fix code style.

* fix yapf

* add cli and demos, fix some code style.

* fix some problem by comment.

* fix yapf

* BytesIO类型时，要保证切到初始位置，这样多次读取才能够正常。比如__call__函数。
__call__函数的参数audio_file为BytesIO类型时执行到self.preprocess(model, audio_file)会报错，需要判断audio_file为BytesIO类型时执行audio_file.seek(0)。