Refactor ETL code to create central utilities

Created by: frankduncan

Right now the ETL pipelines have a lot of redundancy. This should be centralized into some top level library. More than that, that central library should have a more proper object oriented approach that allows for more clarity when doing the kinds of tasks they do. For instance, loading up the spreadsheet into some kind of Datum class that allows you to refer to cells by header name.

This would ideally make the compose-csv parts of the ETL pipelines be much smaller and clearer.