admission: log CPU overload metrics

We log several IO overload metrics through `io_load_listener` to help investigate cases of IO overload. However, we rely only on prometheus exported metrics for CPU. This can create a blindspot when investigating unit test failures that implicate waiting on admission queues.

We should add some elastic CPU and slot metrics to the logs in a similar fashion to the `io_load_listener` logs. Some useful metrics could be:

1. CPU util
2. Elastic CPU util
3. Used/Available slots
4. Runnable go routines
5. Go scheduler latency
6. Token Exhaustion (both elastic and regular)

Jira issue: CRDB-45363

Epic CRDB-55046