Add normalization function for dictionaries of clusters in tests

This will allow for computing whether two lists of `ClusterNode`s are
"equivalent" in the sense that they correspond to the same partition