Non-recursive method to find cells in bounding inverval hierarchy

CUDA devices have problems with recursive algorithms that have no well- defined depth because the stack on a CUDA device tends to be pretty short. Fix the problem for BoundingIntervalHierarchyExec by changing to a state-machine based algorithm that follows the hierarchy up and down.

Also snuck in a name change to prepend CellLocator to the front of the bounding interval hierarchy classes.