Add conversion of state label columns to str if necessary

[l-repa]

pd.read_csv by default sets columns to int or float if they are purely numeric, which causes the string matching done later to fail and return an empty dataframe (which then throws an error when attempting to access its elements)

[ianmkenney]

pandas uses object arrays for strings because of their variable length and is really an issue for performance. I'm not sure how long the object solution will work. It might break if they change string representations. For instance see this experimental feature in pandas.

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.StringDtype.html#pandas.StringDtype

I would instead suggest we tell read_csv which datatypes are needed for each column. You can specify str for the state columns and it will default to the object datatype. I assume that if they change how they represent strings, this will still work.

[l-repa]

Good idea, changed to use dtype kwarg in pd.read_csv instead to specify the state columns should be str.

[ianmkenney]

Thanks @l-repa