Compute labels for a test set according to the k-Nearest Neighbors classification. This is a fast way to do k-Nearest Neighbors classification because the distance matrix -between the features of the observations- is an input to the function rather than being calculated in the function itself every time.