Clark/allow duplicated inputs

[clark-hive]

Currently if all rows of X are the same the proposal distribution q becomes an array of nan's and calling random_state.choice throws ValueError: probabilities contain NaN.

Here's a minimal reproducible example:

import kmc2  
import numpy as np

k=16
x= np.random.rand(8)
X=np.atleast_2d(x).repeat(repeats=10,axis=0)
print(kmc2.kmc2(X, k).astype(np.float32))

I ran into this on the third pytest case in the bolt repo.

This was mentioned later on in the thread where you originally made the fix to the bolt repo: dblalock/bolt#4 (comment) and recently raised again: dblalock/bolt#37.

Or would a better way be to check at the top for the matrix only having 1 unique row and returning that row K times?