Fix MARC records incorrectly matching ISBN promise item records

[hornc]

Closes #9808
Closes #9794

After some investigation, this PR:

• removes the exact match check that was matching existing light records with only title + ISBN against MARC records only on the title field, and skipped the more intelligent threshold match method that at least tries to score matches.
• Fixes a longstanding bug where authors were only being compared on edition level authors, even though in general authors only exist on Work records in OL. This problem was masked by most matches occurring on direct ISBN, or title string only matches.
• Renames the method to find_threshold_match() and makes sure it is used for performing edition matches
• Adds a specific test for MARC imports w/o ISBN should never match light Title + ISBN, undated and no-author bookseller sourced records #9808
• Fixes existing test data which had incorrect author data formats, and did not link editions to works. These issues were masked by the previous title only matching.

Investigation notes

• The threshold_match() test currently passes by not matching the two records.
• The find_match() test currently fails, and I think this is what caused Why did a specific MARC match OL49206154M? #9794

The THRESHOLD checking code seems to do the right thing, but from following through the add_book load code based on your initial investigation in #9794, that threshold checking isn't used for this import flow. If there is a title match, it looks like it counts as an absolute match, and there is also specific code to trigger promise item records getting overwritten by MARC records.

At this stage I'm not even sure that that special case is being triggered here; the record might get added to anyway?

I'm not sure where the 'don't count the match if existing record has an ISBN' code should go. I'm still trying to make sense out of why the threshold code, which appears to be the right thing, isn't being used here.

Feel free to review and let me know if you have any ideas on how to improve this.

Technical

Testing

Screenshot

Stakeholders

@scottbarnes

[hornc]

Following through how and where the threshold match code is used:

• add_book/match.py editions_match() uses threshold_match() (Only 1 usage)
• add_book/init.py find_enriched_match() uses editions_match() (the only use)
• find_enriched_match() is used in add_book/init.py find_match()

in a three stage check:

find_quick_match(rec)
else:
find_exact_match(rec, edition_pool)
else:
find_enriched_match(rec, edition_pool)

openlibrary/openlibrary/catalog/add_book/__init__.py

Lines 838 to 847 in fa22a2e

	def find_match(rec, edition_pool) -> str | None:
	"""Use rec to try to find an existing edition key that matches."""
	match = find_quick_match(rec)
	if not match:
	match = find_exact_match(rec, edition_pool)
	if not match:
	match = find_enriched_match(rec, edition_pool)
	return match

which is effectively:

def find_match(rec: dict, edition_pool: dict) -> str | None:
    return find_quick_match(rec) or \ 
           find_exact_match(rec, edition_pool) or \
           find_enriched_match(rec, edition_pool)

[scottbarnes]

This looks good to me. I also double checked by running the original import from #9794 locally, and the the correct edition is matched with the marc_columbia import.