kvserver: hook up in-memory locks to the memory monitoring infrastructure

**Describe the problem**

Currently, we have a per-range lock limit (which indirectly influences the memory footprint of unreplicated locks). Instead of making this limit per-range, we could instead have a per-node memory limit and have this be hooked up to the memory monitoring infrastructure that already exists. cc @sumeerbhola who suggested this to me yesterday in conversation.

Jira issue: CRDB-55213

No files or tests are named. Start by tracing the existing per-range lock limit and the memory monitoring infrastructure in kvserver, then define how a per-node limit should account for and enforce unreplicated locks; done means the behavior is implemented and covered by appropriate tests.

Written by the indexing model from the issue text.