add preprocess pipelines

[choiHkk]

I found that training with the "VCTK multi-speaker dataset," the training did not work well with the "raw data" In the last code you uploaded, I made some additional changes regarding "preprocess"

The added features are as follows:

text

• Based on version 2.x.x, when using the "phonemizer.phonemize()" function directly, there is a tendency for it to be very slow.
• I added code to call "EspeakBackend" by customizing the backend almost the same as before.
• When using "EspeakBackend," the speed becomes much faster, and you can use it almost the same as before (based on version 2.x.x).

audio

• The "VCTK dataset" contains a lot of "padding" and a considerable amount of "hum noise" in the "padding" areas.
• I attempted to remove it using the "librosa.effects.trim()" function, but it was not easy to adjust the parameters.
• I was inspired by the GitHub repository "https://github.com/nii-yamagishilab/vctk-silence-labels" and added preprocessing code.

[AWAS666]

I also had to use the VAD trim of coqui tts, otherwise the vctk training didnt work at all.

[choiHkk]

@AWAS666 I agree with that. When using a raw waveform that has not undergone preprocessing directly for training, none of the modules related to duration were properly learned, and the loss did not converge either.

[p0p4k]

Thanks for this PR. Can you add a corresponding config.json for vctk with updated cleaners? Thanks a lot!!

[choiHkk]

@p0p4k we just need to change "text_cleaners" from the config file to "english_cleaners3". I've added the config file, so please confirm the config.

[choiHkk]

There is a point I didn't mention. When using preprocess.py to generate filelists, we need to input "english_cleaners3" into the parser due to the following code.

# https://github.com/p0p4k/vits2_pytorch/blob/main/preprocess.py#L17C4-L17C85
...
parser.add_argument("--text_cleaners", nargs="+", default=["english_cleaners2"])
...