Full table scan on MPI_LINK during mdm-clear

[jmarchionatto]

When performing mdm-clear operation for approximately 1,000,000 mdm-links, and associated golden resources, slow sql errors show up in logs as well as other errors associated to DB contention, and the whole operation requires many hours to complete

To Reproduce

1. Generate 1,000,000 mdm links to golden resources
2. Run mdm-clear operation

Expected behaviour
Batch job should advance to a constant pace until finish, without generating slow queries.
Note that is reasonable to expect the operation to run for a few hours to complete for a large number of resources as described, as it runs in batches to allow for other processes to share the server.

HAPI FHIR Version
v6.4.4

[jmarchionatto]

Fixed by #4805