Describe the problem

Transferring leases to unavailable/behind replicas causes outages. In many cases we investigate, when this happens the new leaseholder is waiting for a raft snapshot and until it receives the snapshot will not apply the log up to its lease, but just being behind on a long piece of raft log could essentially have the same effect (though the quota pool should prevent this but probably doesn't always).

When a lease transfer is applied by the recipient of the lease, and the proposed timestamp is (say) >500ms behind "now", the lease recipient should log. This would make it much easier to determine when a lease transfer might have caused a latency spike or period of unavailability.

I believe we could do all of this in this code:

Jira issue: CRDB-23884