Problem:
When the statistics memory limit is hit a log message is emitted. This log
occurs for every transaction. This can cause a flood of messages causing
memory issues, contention, and is not useful at that frequency.

Solution:
Add an internal cluster setting sql.metrcs.discarded_stats_log.interval
to limit how often the log message occurs.

Informs: https://cockroachlabs.atlassian.net/browse/SREOPS-7255

Jira issue: CRDB-31437