Distributed $k$-Clustering for Data with Heavy Noise