EntryIterator should only iterate on data container when local and no loader used

The EntryIterator today does not optimize the times when it is not required to have an off thread queue to guarantee only so many entries are in memory at a time.

The case is when the following 2 things are met:

1. The cache is local only (not distributed) or if CACHE_MODE_LOCAL flag is provided
2. There is no loader to use (or if SKIP_LOADER flag is provided)

This should improve iteration throughput quite a bit in these cases.