Add K-Medians support

[anjali-rgpt]

Proposed feature:

K-Medians is different from K-Medoids, because it primarily uses Manhattan distance and unlike K-Means, it calculates the median of the cluster to form the centroid, thus making it more robust to noise and outliers.
Additionally, while K-Medoids requires that cluster centers are actual samples, K-Medians (like K-Means) can create new data points as centers to minimize the total intra-cluster distance. By default, this is Manhattan distance but it can be done with others as well.

[anjali-rgpt]

@nidhaloff This is another enhancement / contribution which I have resolved with a PR