Off Heap iteration performance improvements

When testing out some things it was found that an off heap cache with a single entry compared to an object one takes significantly longer for state transfer. Off Heap can be improved to make iteration faster and thus improve things like state transfer.

1. Off Heap entries are stored using modulus for locks. This causes iteration to not access entries sequentially as well as having to acquire the same lock many times over.
2. Off Heap should optimize code paths that can be short circuited (ie. don't create stream if size is 0)