Rewrite and generalize post-run evaluation function for all models.

[InfluenceFunctional]

There's now a method for this in the modeller class, where we can write custom evaluation methods on top of standard post-epoch analysis.