Is there any reason this was not built as a pip-installable package? Currently integrating NeMo-speech-data-processor in a data processing pipeline requires a bunch of separate steps to get it installed, and it would be easier if we could just pip install it from Github.