fix: filter with multiple tags

[ThibaudDauce]

The changes on the fix_tags_filters branch add support for filtering by multiple tags (?tag=a&tag=b). Our contribution on top:

Regression test (test_site_csv_exports.py): switching from multi_to_dict to ReuseSearch.as_request_parser() for the /api/1/site/reuses.csv endpoint silently dropped the dataset filter, because ReuseSearch.filters doesn't declare it while ReuseApiParser.parse_filters supports it.

Fix (site/api.py): replaced ReuseApiParser.parse_filters with Reuse.apply_sort_filters() — the same pattern already used by the dataservices CSV endpoint. apply_sort_filters discovers its filters automatically from model field definitions, so there's no risk of desynchronization between two separately maintained filter lists. As a bonus, the generic parser always returns a list for tag values (via action="append"), avoiding the multi_to_dict issue of returning either a string or a list depending on the number of values.

---

Cleanup to do

This bug reveals a broader issue: two independent filtering systems doing the same thing.

The generic system (api_fields.py / @generate_fields): declares filters once on model fields via filterable={}, auto-generates the parser and apply_sort_filters(). Used by v2 endpoints and v1 list endpoints.

The manual ApiParsers (ReuseApiParser, DatasetApiParser, DataserviceApiParser, OrgApiParser): redeclare the same filters by hand in parse_filters methods with chains of if args.get("xxx"). Used by mongo_search() (fallback when the search service is down) and CSV exports.

Both systems support the same filters for the same models, but they're not synchronized — adding a filter in one requires remembering to add it in the other. That's the root cause of the bug we just fixed.

What remains to align:

• CSV exports: datasets.csv, resources.csv and organizations.csv still use DatasetApiParser.parse_filters / OrgApiParser.parse_filters + multi_to_dict. They should use Model.apply_sort_filters() like dataservices and reuses do now.
• mongo_search(): each Search class (ReuseSearch, DataserviceSearch, DatasetSearch, OrgSearch) has a mongo_search() that calls ApiParser.parse_filters + sort + pagination. apply_sort_filters already does all of that. mongo_search could be rewritten to use it, which would make the ApiParsers removable.
• The ApiParsers themselves: once CSV exports and mongo_search are migrated, ReuseApiParser.parse_filters, DataserviceApiParser.parse_filters, DatasetApiParser.parse_filters and OrgApiParser.parse_filters become dead code and can be deleted.