Tools for applying data augmentation to wavs

Since the on the fly data augmentation is not working very well for me and the mozilla voice corpus tool has removed the functionality for varying pitch, echo etc., are there any other good tools available which do the same

Please take a look at this PR: https://github.com/mozilla/DeepSpeech/pull/2897
It should soon get merged.

@Tilman_Kamp Is doing augmentation via the samples likely to be faster than the current SpecAugment method?

@dabinat So far it seems to have almost no impact on training speed at all - given that you keep some of the critical parameters at a reasonable level. In case of caching features to disk it should only cost additional time during first epoch.