restore: enhance scale roachtest with better fingerprinting

One approach to validate that RESTORE ingested the correct rows would be to run either a fingerprint or a row count scan at every backup end time (full and inc) in the backing up cluster, and then rerun the validation command after restore ingests on the restoring cluster. In production, this seems prohibitively expensive-- the whole point of incremental backups is that they avoid full scans via time bound iteration. For revision history backups, this sort of validation is impossible as a user can restore from an arbitrary system time, rather than the finite set of backup end times.

That being said, we could enhance our backup fixture framework to take a fingerprint after the last inc which we will then restore in our roachtests. This would provide additional correctness validation in our roachtest suite.

Jira issue: CRDB-50527

Epic CRDB-50823