codeintel: Do not perform data-retention scans on stale repositories

Created by: efritz

Same effort as #25571. There is an obvious bug obviously a bug that causes valid records to be deleted while the repository's commit graph is stale. I noticed this on the codeintel-qa pipeline, which always starts with a fresh instance. We were not seeing this on local instances which were already running because the test timing happened to miss the data retention janitor. Ensuring they run concurrently shows a lot of records marked as expired that shouldn't be.

This PR changes the data retention process to only work on repositories with updated commit graphs. This is a simplifying condition, and the only change in behavior here is the frequency that records are removed for repositories that are always stale. This is another issue entirely.