SSE v1.0

Adds

• new training code supporting distributed training, fp16, logging, and more!
• new pretrained model for 16kHz audio that operates directly on raw waveforms.
• benchmarks of the new and old model in terms of EER metrics for Librispeech test dataset
• updated readme and usage notes
• bug fixes

Initial model release

0.1

the big fix