Task: groupby (i.e. "split")

[cisaacstern]

This task uses pandas.Dataframe.groupby internally to apply the groupers defined by #72, and split an input dataframe into groups for processing.

It may be easiest to do this together with #72 since these features are closely linked.

Depending on the setting (local, cloud) and/or workflow configuration, this task may emit either actual dataframe groups, or possibly references (urls) to serialized groups (i.e. hive-partitioned parquet). The latter is demonstrated in #45.

We should also leave room for the eventual scenario in which the data may be live in a backed that supports SQL queries (BigQuery, DuckDB, etc.), in which case the groupby may not need to happen in-memory. This is not relevant to our immediate concerns but, if there's a way to do so that doesn't take much extra time, it may be useful to make a nod to this idea somewhere in this implementation (a NotImplemented code path, etc.) so we keep it in mind.

[cisaacstern]

closed by #162