fix(harvester): handle legacy CDS recid checks before create

[TahaKhan998]

Closes #907
This PR changes how the INSPIRE harvester handles records with a legacy CDS recid and avoids duplicates between old CDS and CDS-RDM. If the incoming record has an integer legacy recid, we first check pidstore and update when it is found. If it is not found, we check old CDS directly. We raise an error when it redirects to repository.cern or returns 404. We skip with a warning when the record is still only in legacy. If there is no legacy recid in the incoming record, the existing matching flow stays the same. If INSPIRE sends more than one CDS recid (merged records keep both), we check each in pidstore and on old CDS. Two different pidstore hits are treated as multiple records match.

[kpsherva]

I am not fully sure why this is removed from expected result - the results should still have CDS legacy recid inside the metadata, if this not how the records are created in the end then it might be incorrect - if record had legacy id and is ingested, it should be harvested with the legacy id in the metadata

[TahaKhan998]

we still map the cds recid into metadata.identifiers when it’s on the inspire record, this test is a create test though and after this pr if the fixture still has a numeric cds recid the matcher thinks it’s still on old cds and skips so nothing gets created and the test fails, that’s why we strip them from the mock inspire payload with drop_legacy_cds_ids and why they’re gone from expected_result, the input doesn’t have them anymore. migrated records still keep the legacy id, that’s what add_legacy_recid is for in the update tests.