* Implemented data augmentation for concatenation  (#3516)

* Implemented ConcatAug as setting from config

* Switched ConcatAug implementation to sweep script

* Added rate and max tokens as ConcatAug params

* Kept original fns, pulled concat_attempts as hyperparam

* Fixed ConcatAug nits

* ConcatAug typing recognizes int and np.int

* Implemented waveform transforms and suite of noise augmentation techniques (#3517)

* Implemented ConcatAug as setting from config

* Switched ConcatAug implementation to sweep script

* Kept original fns, pulled concat_attempts as hyperparam

* Implemented WaveformTransforms, MusicAug

* Removed leftovers from debugging

* Separated out feature_ and waveform_transforms, updated constants, formatting cleanup

* Added Babble and SporadicNoise augmentations

* Fixed zero division error

* Adding BackgroundNoiseAugment

* Added warning for if using feature transforms with waveform input

* warnings, SNR fix

* fix for NoneType extension error

* fix 2 for NoneType extension error

* delete print

* Dataset transform, NoisyOverlapAugment, reframe ConcatAugment (#3533)

* Dataset transform, NoisyOverlapAugment, reframe ConcatAugment

* using np.random instead of python random

* fixed np random upper bound bug

* cleanup

* Changed args & return expressions for waveform transform

* Documented new augmentation features

* Create augmentation_example.md

* Update augmentation_example.md

* Update, benchmarking left to do

* Move docs to speech_to_speech

* Remove docs from speech_to_text

* [docs] Updated clean benchmarks

* [docs] Add benchmark data