schemachanger: add UNDROP TABLE

Currently, if a user accidentally drops some tables, they can recover via RESTORE or CREATE TABLE AS OF SYSTEM TIME. We could build an alternative approach that doesn't require a backfill: via an UNDROP table cmd that can be run before the gc TTL. For this to work, we'd likely need to refactor DROP to:
- provide some unwinding of all the references dropped [here ](https://github.com/msbutler/cockroach/blob/butler-read-only-fail-upgrade/pkg/sql/drop_table.go#L244)
- delay the dropping of the del range until the ttl for the table has expired by polling its span configs

So, if the user runs UNDROP before we issue a delrange, we could, in theory, skip the del range drop and rebuild the schema graph.

I don't know if this is technically sound, nor do I think this should be prioritized. I'd rather just make restore or index backfills very very fast. But obviously i'm biased and not a PM.

cc @arulajmani

Jira issue: CRDB-49876

Epic CRDB-61003