Include the possibility of performing distributed learning of AMIDST probabilistic models on a SPARK cluster.

The branch sparklink already contains some functionalities for reading data, sampling from a Bayesian network and performing Maximum likelihood estimates.