kvserver: use count of encountered MVCC tombstones as GC signal

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

See https://github.com/cockroachdb/cockroach/issues/17229#issuecomment-1460942252. In this issue, @a-robinson ended up with a range that wasn't eligible for GC even though most of the shadowed data was GC'able, because the GC'able data represented only a small portion of the range. However, this small portion of the range was hot and so performance suffered.

**Describe the solution you'd like**

We could capture stats the rate at which shadowed versions are skipped and use this to influence the aggressiveness of GC. In particular, the `DeadFraction`[^1] should not lower the score when the shadowed data is read at a high rate.

[^1]: https://github.com/cockroachdb/cockroach/blob/736a67e0d36cc545bf74d65db069ee895ff9bea0/pkg/kv/kvserver/mvcc_gc_queue.go#L457

Jira issue: CRDB-25331

Start with pkg/kv/kvserver/mvcc_gc_queue.go, especially the DeadFraction reference, and read the linked discussion in issue 17229. Determine how encountered MVCC tombstones and the rate of skipped shadowed versions should influence GC aggressiveness, then verify that hot shadowed data can raise the GC signal.

Written by the indexing model from the issue text.