Fix #35. Ensure parcel_count attribute is persisted to CSV output.

[parkerziegler]

This PR fixes the small issue outlined in the first comment of #35. Importantly, it does not remove the SKIP_DEDUP flag for MN (though that should be trivial once this is merged).

The Problem

This line was attempting to write an attribute "PARCEL_COUNT" (note the all caps) to each GeoDataFrame. Tracing program execution from this line, the dataframe (after some transformation) is handed off to _merge_dataframes, which in turn calls combine_dfs for cases where we need to merge more than one dataframe.

Inside of combine_dfs, we do some set computation to build up the set of attributes shared by all dataframes while also ensuring columns present in the constant list FINAL_DATAFRAME_COLUMNS are also present. Later, on this line, we subsequently drop all columns not in FINAL_DATAFRAME_COLUMNS. The attribute "PARCEL_COUNT" is not in that list; instead, the string literal "parcel_count" (note the lowercase) is. This literal is represented by the constant PARCEL_COUNT.

The Solution

Pretty small—just change the string literal "PARCEL_COUNT" to the constant binding PARCEL_COUNT, which represents the string literal "parcel_count".

Along the way, I did notice an opportunity to remove a tiny bit of dead code from overlap.py. We already return early in combine_dfs when we have a df_list of length 1, so there's no need for the conditional check on df_list length afterwards.

[clayton-aldern]

Bless you @parkerziegler; I was ready to jump off a bridge.

[parkerziegler]

It took a good while to see it, definitely had me questioning reality for a bit 😂