Extract non-glue code from notebooks

[amy-langley]

Right now there's a lot of code that is copied and pasted between notebooks and/or notebooks that define a lot of functions and then call them. Function definitions should be moved into .py files so that they can be reused easily and automated with snakemake.

• [x] encapsulate legiscan api
• [x] NLP functions
• [x] tokenize corpus task
• [x] retrieve legiscan datasets task
• [x] extract legiscan datasets task
• [x] process legiscan datasets task
• [x] prepare neutral corpus task
• [x] generate stopwords task
• [x] update aclu dataset task
• [x] update ttl dataset task
• [x] legiscan lookup task
• [x] document download task
• [x] metadata download task
• [x] aggregate task
• [x] categorize task

[amy-langley]

There's probably even more to be done but this is great progress.