Filter out invalid CSV values so they never get ingested

[mjordan]

Now that #417 is done, any incoming values that are invalid ETDF, link, geocordinates, etc. should be filtered out. Related issue #387.

[mjordan]

Work started in f45d49e on SimpleField, GeolocationField, and LinkField classes, with tests in field_tests_values.py. Need to figure out how remove_invalid_values() will work with 1-cardinality creates/updates.

Validating entity reference and typed relation values will likely need a hot Drupal, so we'll need to figure out how to configure host, username, and password in tests.

[mjordan]

Note about the status of this work added to https://mjordan.github.io/islandora_workbench_docs/preparing_data/#how-workbench-cleans-your-input-data.

[mjordan]

Reviewing this feature again. For every field type, we want to, in this order:

1. remove invalid values. We never alter values (e.g. truncate text longer than allowed) except for title; instead, we do not populate the field (or value instance in multivalued fields) and just log that we haven't populated the field.
2. dedupe values