TODO: Implement heuristic classification of system based on observables

[DM-Berger]

Once #4 is resolved, and under the assumption of accurate unfolding, then there should be a way to roughly classify a set of unfolded eigenvalues as being, for each observable, either "more GOE" or "more Poisson / GDE" based on their comparison with the expected curves for each observable. This should be implemented, probably as a method on Unfolded class:

class Unfolded(EigVals):
    def ensemble_score(self, observables=["nnsd", "rigidity", "levelvariance"]) -> Tuple[Float, Float]:
        return goe_compare, poisson_compare

However, it would probably be good to also have a way to ensure the comparison is not overly sensitive to the choice of unfolding method, and that there is some way to identify curves that are neither clearly GOE nor Poisson (e.g. for mixed systems).