Saddle points and borders in clustering.py

[charliematteo]

1. current implementation refers as "_bord" to attributes concerning only the saddle points and not the whole border between two clusters
2. like in the array "cluster_indices" we should store a matrix "saddle_point_indices" or something like this
3. we might also save a matrix of lists "bord_indices[i,j]" which at any entry [i,j] contains the list of indices of points in cluster i at the border with cluster j