velocity cluster finding for finding primitive gamma bands

[ExpHP]

a possible solution to #33 is to put aside general algorithms for a moment and focus on what seems to be a property of these bands:

Given a band, consider the distribution of all the per-atom velocity vectors as a collection of points in 3-dimensional velocity space. For acoustic modes, pretty much all of these velocities should be the same, so we'll see a single dense clump of points. For layer acoustic modes, we will see up to two "clusters" of points; one for each layer.

It seems to me that one could reliably expect that, for a material with L layers, then the modes that correspond to the primitive cell gamma are the 6*L modes with the fewest number of velocity clusters.

[ExpHP]

Professor wants bands at K now.

Probably better to just do unfolding within each layer separately