Almost all pools allocated 2 chunks on initialization, every element needed to be added to the free-list which would never be used for small pools. Now allocate only one, gives minor speedup for some bmesh operations.