fixed bug related to non-constant number of peaks found with the blob finder

The bug was due to the order at which the peak-ellipsoid pointers were added to the Octree for finding putative collisions. To solve this, I introduced an intermediate step where the ellipsoid pointers are added to a vector which is further sorted by increasing xyz of the pointed ellipsoids. The elements of this vector are then added to the Octree which ,in doing so, hosts the ellipsoids in a consistent order. Tested successfully on 100 runs of TestNewWorkFlow which all found 854 peaks.