Fix a bug with the override_column_X attributes in conf_validations.py

[riley-harper]

Fixes #130.

In issue #118 we documented the override_column_a and override_column_b attributes for column mappings and added some tests for those. I forgot to check the config validation though, and there was no logic in there for handling the override_column_X attributes. So if you used the override_column_X attributes in a column mapping, you'd still get a config validation error from the conf_validations module. If you decided to run hlink anyways, it probably ran without errors. The bug was just in the analysis code, not in the core module code.

I've added some more unit tests for this section of the code and refactored the check_column_mappings() function to make it easier to work with.

The changes to hlink/linking/matching/link_step_explode.py are black formatting changes. black has updated since we last made a change to hlink.

[joegrover]

It's crazy how readable this is while at the same time being so confusing. The logic all tracks, though I'm not entirely sure what it is trying to do. Or at least it seems like it is doing more than what I initially assumed was a simple "call COLA from file A COLB in file B". Specifically the concept of a "previous alias" caught me off guard. Do these things stack?

[riley-harper]

Haha, yeah this is something that feels very simple at first but is more complicated than it seems. There is some documentation for column mappings here that might be helpful.

Yes, the previous aliases do stack. So if you have a column mapping in your file that reads in a column as AGE, you can have another column mapping later that uses the column AGE as input. In hlink/linking/preprocessing/link_step_prep_dataframes.py line 106 the column mappings are selected one at a time.

[joegrover]

Ah, got it. Yeah that makes sense then.