Rename .sty field on TyS

[varkor]

.kind makes more sense, but could be confusing, as it's not a type-theoretic kind. There's also the issue of having inconsistent ways to name a field containing a *Kind, e.g. .node and .kind. It would be good to make these consistent.

[Centril]

I don't like ty.kind -- that's a bit of high-degree confusion in the heart of the type system.

I'd like to suggest a third alternative .form and ExprForm to use consistently everywhere.

[varkor]

@eddyb suggested .which.

[Centril]

I would like to have .thing and ExprThing be consistent. Idk if ExprWhich works so well here.

Another alternative: .shape: ExprShape.

[petrochenkov]

I'm for FooKind / foo.kind, which is an existing and widely used convention.
TyKind / ty.kind is also ok.
The confusion with type-theoretic kinds applies only to a couple of selected cases, and even in case of ty.kind it's clear what kind it is, not worth changing conventions everywhere.

.node is a legacy convention, I always thought is didn't make much sense, personally, I'd gladly replace it with .kind in AST/HIR structures.

[oli-obk]

Alternative suggestion:

get rid of the fields on TyS and make the typeflags a query as well as the outer_exclusive_binder: DebruijnIndex something that is "computed" on the fly (most types are just innermost, but the few that aren't can have this info in their variant)