When doing a rolling restart we recommend disabling shard allocation to
avoid unnecessary recoveries. However, this advise is unnecessary or
even harmful when restarting nodes that do not carry any data like a
pure ML node.
Synced flush was a brilliant idea. It supports instant recoveries with a
quite small implementation. However, with the presence of sequence
numbers and retention leases, it is no longer needed. This change
removes it from 8.0.
Relates #5077
