storage: add a more convenient sql function for compacting a table

We have `crdb_internal.compact_engine_span` to compact a span, but it is very low-level: you have to specify the exact node ID, store ID, and raw keys. It's hard to manually get this information.

We should implement a higher-level function that takes a table (optionally with an index) and compacts the relevant ranges (one at a time, to minimize disruption). This would be very useful for escalations.

Jira issue: CRDB-29059