Goal: 

currently, all dataset processing related functionality lives in `token_dataset` and `data_process` in the training library. These modules should be further split up into as many individuals units as possible. Allowing for users to pick and choose specifically which data processing functionality they want to utilize from the training library.

Also consider refactoring the class structure, mode of access (argparse cli vs importable classes, click cli) in this process.

Acceptance Criteria:

Training library data processing has been re-factored into multiple individual modules that each have a distinct use and purpose.