Distributed Stream rehash can miss entries on non replicated caches with non iterator methods

When fixing some rehash test failures in ISPN-9542, we added in an optimization to not allocate a listener and additional objects when using a replicated cache. Unfortunately in doing so the listener is only registered with replicated caches on accident. This leaves a small window between when we start iteration (we check segment) and the completion that if we lose an entry due to state transfer.

The window is small enough to hit that it almost never shows in DistributedStreamRehashTest, due to it requiring state transfer to begin and complete between the start and and completion of iteration, which is a hard task when it actually has to move entries.