storage: additional timeseries metrics for improved out of disk debugging

**Is your feature request related to a problem? Please describe.**

It is often useful when debugging disk space issues to be able to deduce whether Pebble is a likely culprit - e.g. large in-progress compaction pinning tables in the LSM for the duration of the compaction; zombie tables / memtable leaks.

**Describe the solution you'd like**

Add at the following timeseries metrics that can be used in a custom chart:

- In progress compactions - size and count
- Zombie tables - size and count
- Zombie memtables - size and count

**Describe alternatives you've considered**

Use the LSM printout to deduce. The granularity isn't great here, as the stats are only printed in the logs every 10 minutes.

**Additional context**

More context in [this support case](https://github.com/cockroachlabs/support/issues/2092#issuecomment-1443989245).

Jira issue: CRDB-24893