removed some cruft from Superpose3D()

It turns out that dividing the coordinates by the size of each point cloud (Rgf and Rgm) before superposition does not make the algorithm more numerically stable after all. So I removed the unnecessary and confusing code related to this bad idea.