Computing distances twice for random landmarking

[MattScicluna]

I think we can avoid this computation (happening during landmark creation):

graphtools/graphtools/graphs.py

Line 1066 in c18d494

distances = cdist(data, data[landmark_indices], metric="euclidean")

Since we compute the distances between every pair of points here:

graphtools/graphtools/graphs.py

Line 698 in c18d494

distances, indices = knn_tree.kneighbors(Y, n_neighbors=search_knn)