kvserver: re-evaluate default merge queue interval cluster setting

**Describe the problem**

Because of the expense of merges, the default merge queue interval is 5 seconds (e668bfe8d21556d8970456e146e4000768faea03). As a result, a given store only processes about 12 replicas per minute. As a result, once the number of replica's exceeds the queue max, the merge queue is perpetually full and it can take a considerable amount of time before a given range is actually processed by the merge queue.

In clusters that have a high rate of producing many small ranges, the cap on the merge queue can result in an ever growing range counts. This is sometimes caused by workloads that do a large number of scheme changes or workloads that consistently truncate tables.

Even in clusters without those workloads, we occasionally find that lower the merge queue interval for a brief period of time results in a substantial range count reduction because of small ranges that scanner was simply unable to ever add to the queue because it was full.

Jira issue: CRDB-63016