Clustering Linear Models Using Wasserstein Distance