Recursive priorities #55
Description
If we wanted to quash all priority issues, once and for all, we'd have to go one final, additional step from #54: rather than constructing Schema objects lazily, in validate(), we should traverse all the way down the hierarchy, turning things into Schema subclasses, in Schema.__init__ (or, more likely, in Schema.__new__ or some other factory function, so we can spit out different classes). Then we would have the complete compound priority, like (VALIDATOR, VALIDATOR, COMPARABLE) for Optional(And(int, is_even)), and any remaining problems with nondeterminism would be scared into a little corner where we can't decide which of, say, 2 callables has higher priority. If someone wanted to explicitly set priorities, they could write their own class with a priority() method, or we could even introduce a manual priority= kwarg on Schema for those special, hard-to-reach cases—or for cases where people don't want to go by our concept of specificity.
Creating this issue as a place for discussion about this.