Faster type equivalence testing

[shirok]

Simple types and aggregate types with the same constructor arguments can be compared with eq? so it's fast, but there are cases that non-eq? types are equivalent, e.g. (</> <integer> <string>) and (</> <string> <integer>). We can test equivalence with (and (subtype? X Y) (subtype? Y X)) but it can be slow.

Adding equivalence test callback to <type-constructor-meta> may help testing equivalence fast.

Internally we can sort the elements of choices (the order shouldn't matter as type), so that we can compare them faster.

[shirok]

Sorting types isn't that straightforward. We couldn't comapre pointers of <type> instances, for there can be more than one instances of a type. We need to devise a consistent ordering including aggregate types.