Load testing of user mapping for Direct Transfer (required before releasing the feature)

After User mapping - Integration with Direct Transfer (#443557 - closed) we should load test the performance of user mapping to proactively discover performance/resource contention issues.

We should do this before we release the feature.

On a test server (staging, and/or our 3k instance):

• Conduct several massive Direct Transfer imports simultaneously to see the performance of the creation of placeholder users and saving of contribution references to the table created in !156241 (merged). The imports should be on the scale tested in https://gitlab.com/gitlab-org/gitlab/-/issues/429981 but also with 1000s of unique users.
• Trigger many assignments of placeholder contributions simultaneously to see the performance of the reassignment process #443556 (closed). Each placeholder should have 10,000+ contributions that need to be assigned.

We want to verify that:

• All the data is created successfully
• Load on servers is acceptable (to be defined, but we have https://gitlab.com/gitlab-org/gitlab/-/issues/429981 previous example)
• There are no/few resource contention errors
• ...

Areas to assess in particular:

• Look up and creation of placeholder users during import scales
• Saving placeholder user references scales
• Updating of records during the reassignment process scales

And see if there are any meaningful places to:

• Optimise the work being done to be more performant (to do it faster, or with lower resource consumption)
• Apply better logging
• ...