This enables us a fast reference to the entities current chunk instead of having to look it up by hashmap lookups. We also store counts by type to further enable other performance optimizations in later patches.
