We know operations like findOneByID/Index, updateOneByID/Index, insertOne, deleteOneById/Index should have very low latency. We should create histograms showing how long they actually take to show how stalls can affect user operations beyond the canary.

SERVER-127567 already added queryLatencies.express to cover the read queries.

The goal is to provide health metrics for the host itself, so we should exclude time spent waiting for write concern (on ASC) or for a log server (on DSC).