This PR de-duplicates, using `casefold()`, the `subjects` field on
`Work` items, and field values of list items that are added to `Edition`
items on import via `load()`.

This does not affect edits made via the UI as they go through
`process_work()` from `openlibrary/plugins/upstream/addbook.py`, which
de-duplicates via a slightly differest strategy. Rather than merging
two lists (existing subjects on a matched item, new ones from an import
item), it takes form data with every subject and dedupes that list.

Though I had hoped to unify the de-duping logic, I think that is beyond
the scope of this particular issue.

For more, see:
internetarchive#8661

- fix comparison
- remove pointless sorting
- ensure tests catch case where 1 existing duplicate is removed and 1
  new item is added, resulting in a final list the same length as the
  original.