storage: disable crashing for single delete callbacks

As elaborated in code comments, these can be false positives caused by
delete-only compactions. The metrics are not disabled, under the rare
chance that we see a problem in a cluster and one of the metrics (if
it happens to not be a false positive) gives us a hint about the cause
of the problem. For the same reason, we log such callbacks every 5
minutes, so we can correlate the key in such logs with an actual
problem. False positives should be rare, especially for the invariant
violation callback.

Informs #115881
Informs #114421

Epic: none

Release note (ops change): The cluster settings
storage.single_delete.crash_on_invariant_violation.enabled and
storage.single_delete.crash_on_ineffectual.enabled are disabled and
must not be enabled.