Distributed Learning of Finite Gaussian Mixtures