adding ability to add multiple values of k

[jkobject]

to merge after fix_for_numba_jit.

ability to choose different ks for each, can really improve the algorithm's ability to perform mnn when we have very diffferent dataset sizes. like 10 to 1.