RFE: Decaying locks

[ananace]

One type of lock that would be useful would be a decaying lock, one where a lock automatically decays into an open state after - a specifiable - length of time, so that machines have to actively poll the lock in order to be allowed to keep it.

Additionally, as an extension on top of #1, having a decaying lock where only N machines are allowed to release the lock at the same time would be very useful when wanting to make sure that a cluster only allows a certain amount of machines to be unavailable at the same time - be it due to scheduled or unscheduled reasons.