Check that the MLE DG estimates do not depend of the order in which DDG values are fed in

[hannahbaumann]

The doc string of the MLE states that We assume the free energy of node 0 is zero.. It should matter which node is set to zero, but we want to double check this since one of our partners showed some results where the same dataset gave different results using different networks. Here we only want to doublecheck that results are not just shifted by some value.