Enable the user to save the preprocessed wavs

[MENGZHEGENG]

Sometimes the user may be curious about what the audio will be like after our preprocessing steps.

We have an API to enable the user to save preprocessed waves, which is good:
https://github.com/roedoejet/EveryVoice/blob/d117ece5d29947e171bf181bc481c66bb26352ff/everyvoice/preprocessor/preprocessor.py#L121C9-L121C18

I'm sorry that I didn't figure out how to configure save_wave to TRUE in the configuration files.

[roedoejet]

I think we should just only save .wav files, and not .pt. but we can do a benchmark for training speed between .pt and .wav.