Add data augmentation scripts to work natively in PyHa

[sprestrelski]

Add ability for the data augmentation scripts that can filter and scale audio, as described in this paper and developed by previous members of the project, to work natively in PyHa.

[JacobGlennAyers]

All of those scripts can be found in this Microfaune training pipeline: https://github.com/UCSD-E4E/AID_ICML_2021/blob/master/Train_Microfaune_RNN.py

With that being said, I feel that we would need to first introduce the scripts to retrain the relevant neural networks used (TweetyNET, Microfaune, and maybe BirdNET in the future if they release it) because I don't see a purpose of data augmentation without those pipelines being available.

[sprestrelski]

That makes sense. The REU2022 team is working on retraining pipelines for OpenSoundscape, TweetyNet, and BirdNET-Analyzer with the idea that we can integrate them into PyHa later. Microfaune was put on the backburner because of a TensorFlow 1.0 -> 2.0 compatibility issue

[sprestrelski]

I don't think this fits within the scope of PyHa and should instead be in https://github.com/UCSD-E4E/acoustic-multiclass-training