Cache node counts using SQL as key

In some cases, eg going back and forth between node configuration, we could benefit from caching node count SQL

Need to be able to turn off though, for benchmarks. Some things eg classifications etc may not be able to be cached

No files or tests are named. Start by locating the node-count SQL call sites and existing cache configuration, then inspect how benchmarks could disable caching and which cases, such as classifications, must remain uncached. Done means the safe cache scope, SQL key behavior, and opt-out requirement are defined and covered by appropriate tests.

Written by the indexing model from the issue text.