Follow-up from "Dedup incremental checkpoint blobs for redundant delta sends"

The following discussion from !243339 (merged) should be addressed:

• @stomlinson started a discussion: (+1 comment)

  Not a blocker for this MR, but will this bind the created_ats together so that we can restrict the created_at range when we query a CheckpointBlob? In general, we really want all queries to a table to include a partitioning key, so that we only scan a single partition per query. Otherwise it gets a lot more expensive for postgres, and we can end up with problems with LockManager LWLock contention at high call rates.