[BUG] make_blobs doesn't behave like cuML and scikit-learn counterparts, and shuffle doesn't really shuffle

[Nyrio]

This is better understood by example.

cuML and scikit-learn behavior:

unshuffled: [0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5,5,6,6,7,7,8,8,9,9]
shuffled:   [0,4,9,4,9,8,5,6,3,6,0,8,4,3,2,7,1,5,0,7,1,2,2,3,1]

raft behavior:

unshuffled: [0,1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8,9,0,1,2,3,4]
shuffled:   [8,5,2,9,6,3,0,7,4,1,8,5,2,9,6,3,0,7,4,1,8,5,2,9,6]

The difference between the unshuffled versions is cosmetic (cuML and sci-kit learn have continuous labels, raft has the index modulo the number of labels). However, the "shuffled" version in raft is not properly shuffled, as the labels appear cyclically in the same order.

This is due to how raft attempts to shuffle them, using an affine transform. Modulo congruence is not sensitive to multiplication and addition. That is, if two points have the same label pre-transform, they have the same label post-transform.

We should use a different method to shuffle.