Two distributions are `==` iff they have the same `support` (but not necessarily same pool) and probabilities on that support agree.