performance: SQL throughput and latency impact during rebalancing

**Describe the problem**

During a recent performance investigation, we a zone configuration change resulted in rebalancing which substantially impacted foreground SQL latency and throughput.

The rebalancing period:

The impact on foreground throughput and latency:

Throughout this period, CPU usage and other metrics that typically indicate overload all looked good. The rate of NLHE looks relatively low compared to the request rate:

Jira issue: CRDB-61859