Identify exact matches pre-dedupe

[TaiWilkin]

Overview

When matching facilities, look for exact matches prior to entering
the dedupe process. Exact matches (for name, address, and country)
can be automatically matched, and don't need to enter the dedupe
process.

Adds new fields to FacilityListItems to store the cleaned name and cleaned address, allowing efficient filtering for exact matches.

If multiple exact matches exist, we prefer first the active list items; then items submitted by the same contributor; then finally the most recent items.

Connects #1549

Demo

Notes

PPE handling was not implemented for exact-match items.

Testing Instructions

• Run migrations.
• Login as c2@example.com, and navigate to http://localhost:8081/api/docs/#!/facilities/facilities_create. Submit an exact copy of an existing facility with create=true:

{
"name": "Alif Embroidery Village Ltd.",
"address": "Bangobandhu Road, Tangabari, Ashulia, Savar, Dhaka 1341",
"country": "Bangladesh"
}

• Confirm that the item has matched. Note the facility id. In an incognito browser, login as c1@example.com and navigate to the new FacilityListItem in the Django Admin. In the processing results, you should see exact_match: true.
• As c1@example.com, navigate to adjust facility matches. Transfer the match you just created from the facility it matched to to any other facility.
• Resubmit the same facility via the Swagger API. It should now match to the facility you transferred the FacilityListItem to.
• Deactivate both newly created matches. Submit the same facility details via CSV.
  duplicates_test.csv
• Process the list:

./scripts/manage batch_process --list-id 16 --action parse
./scripts/manage batch_process --list-id 16 --action geocode
./scripts/manage batch_process --list-id 16 --action match

• Confirm that the newly uploaded item matched to the original facility via exact match.

Checklist

• fixup! commits have been squashed
• CI passes after rebase
• CHANGELOG.md updated with summary of features or fixes, following Keep a Changelog guidelines

[jwalgran]

The implementation looks good and I was able to run through the test successfully. We will want to pay attention to how this performs on staging when we have 2 orders of magnitude more list items.

[jwalgran]

After running through the testing instructions, the last list item ended up with 3 match rows. I originally thought this was a problem, but I confirmed that this matches the behavior of the normal deduple process where we would create matches for both a 60% and 90% confidence match but the high quality match would be the AUTOMATIC winner. 👍

openapparelregistry=# select id, results, status, facility_id, facility_list_item_id from api_facilitymatch where facility_list_item_id = 935;
 id  |                 results                  |  status   |   facility_id   | facility_list_item_id
-----+------------------------------------------+-----------+-----------------+-----------------------
 937 | {"match_type": "multiple_exact_matches"} | AUTOMATIC | BD202200698XDZC |                   935
 938 | {}                                       | PENDING   | US202200661FFPF |                   935
 939 | {}                                       | PENDING   | US202200661FFPF |                   935
(3 rows)

[TaiWilkin]

Thank you for reviewing!