Consider refactoring post-validated block types

[sco1]

The typing is a mess. Currently, typing for serialization uses the types defined by the parser which are broader than what comes out of the validator. This leads to a lot of type narrowing that isn't necessary because the validation step has thrown out values that aren't acceptable for each parameter.

I think it's possible to narrow the post-validation types into something structural, so let's look at that soon.