Add the impact of the reform by county

Add three raw files:

• data/raw/[shapefiles for puma, assembly, and senate]

And two files for processing:

• py/make_puma_mapping.py: Python script
• data/puma_mapping.csv: File with puma, assembly_district, senate_district, produced from make_puma_mapping.py

For California for now since others require separate assembly and senate files.

Currently @nmrodelo has some matplotlib charts, we'll want plotly so we can have hovercards based on @ben-boehlert's outputs

Nationwide to start

Since state legislative districts are built from Census blocks, we should build a dataset at the block level that we can then aggregate as needed. It should have these columns:

• Block
• Block group
• Tract
• Upper state legislative district
• Lower state legislative district
• PUMA
• State
• Population*

* Population can probably be mapped from the 2010 Census, which has block group level population; this can then be allocated equally across blocks in a block group.

censusdata package should do this

Depends on #1 and #2.

With links to the sources of those files (shapefiles and ACS-SPM data)

Since it's unicameral.

Should also test that records don't get dropped in general.

I asked about this on StackExchange since I can't find it.

Somehow the centroid match is generating errors for some of the districts, and providing text instead of a number. It is not immediately clear how this is happening, but it may simply be the result of the PUMA with centroid off the coast.