Ellipsoid factor averaging over several runs should re-use resources

[mdoube]

Describe the bug
Ellipsoid Factor can calculate an average over several runs to help smooth the result. The current implementation is a bit inefficient on memory because several image copies are held in RAM and inefficient on CPU because seed points are recalculated on each iteration.

Expected behavior
Re-use of already calculated values. Not making data copies unnecessarily.

P.S. ID image has channel and z-axes swapped, possibly true for other images.