Add test for GaussianMixture with distributed decompositions

test("check distributed decomposition") {
    val k = 5
    val d = decompositionData.head.size
    assert(GaussianMixture.shouldDistributeGaussians(k, d))

    val gmm = new GaussianMixture().setK(k).setSeed(seed).fit(decompositionDataset)
    assert(gmm.getK === k)
}

In ML GaussianMixtureSuite, the above test only check that when we distribute the computation that it produces a model, but does not check that it produces a correct model.
It should have, but depends on GaussianMixture can be set with initialModel (SPARK-15785), otherwise, this algorithm seems incapable of learning even this very contrived example. After that was resolved, we can add a correctness test in the above case.