Figure out what to do about the backlog of MARC21 + MARCXML-formatted records used for testing

[cbeer]

We're carrying around a lot of MARC21 + MARCXML data used in the tests converted from solrmarc. We're moving away from MARC21 + MARCXML in our indexer (FOLIO uses a JSON representation internally), so it's probably time to do something about this data, e.g.

• build on-demand MARC data (or Folio record instances) for unit testing specific scenarios
• be a little smarter about generating MARC data used for the shotgun "what happens if we map X" tests
• rewrite tests to test specific records instead of "run the file through the indexer and pluck out particular documents" approach
• convert remaining MARC21/MARCXML to the MARC JSON schema FOLIO uses

[jcoyne]

There are 1101 instances of select_by_id in the tests where the "pluck out particular documents" approach is used.