This issue builds on top of persisted statistics by enriching the DB console UX. Rather than the last hour, we will now surface persisted statistics which has a default 1-hour aggregation interval (configurable). We round to the nearest hour for each interval (if the cluster starts at 6:16, the first aggregation window is 6:00). The retention period is sql.stats.cleanup.max_row_limit where the initial default is 10K. High-level improvements:

1. The transactions and statements pages should be updated to accommodate persisted statistics
2. Persisted statistics also surface changes in the query plan (via a hash and JSON plan tree). We should accommodate this experience in the statements page for query plan regression detection.
3. Given a longer retention period, we should see if/how we can now correlate the time-series SQL metrics with the statements and transactions pages

V1:

1. The user can specify a time range (start and end time) that will surface all intervals (and their fingerprints) within the specified range on the statements and transactions pages.
2. The user can reset stats which will clear/truncate both in-memory and persisted stats.

Epic: CRDB-394