We have observed that sometimes SQL Stats cleanup job runs way too slow. (sometimes 12 hours+). This can be problematic when the are sql stats flush rate is high. We need to find a way to speed up SQL Stats cleanup job.

cc: @ajwerner

Update: action items: (cc @mgartner )

• Knob to tune the 128 row limit.
  • sql: add cluster setting to tune sql stats cleanup job behavior #80191
• Update the query to skip over garbage.
  • sql: rewrite sql stats compaction job to avoid scanning mvcc garbage #80341
• Add timestamp filter to query to prevent deleting too many rows.
  • sql: ensure SQL Stats cleanup job to not interfere with foreground traffic #80515
• Remove the initial count query and instead have the for loop delete queries until < 128 (or the setting value) are deleted.
  • Partially addressed by sql: ensure SQL Stats cleanup job to not interfere with foreground traffic #80515
• Consider increasing the number of shards in the two tables in future versions.

Jira issue: CRDB-14902