Consider precedence btw pat : type (ascription) & p0 | ... | pn (or_patterns)

[Centril]

We should ensure that we consider the precedence of type ascription in relation to or-patterns, e.g.

let A(x) | B(y): T;

should probably bind as (A(x)) | (B(y): T) and not (A(x) | B(x)): T. This might not make a semantic difference because both sides should be of the same type, but parsing in the right way is a more iron-clad guarantee of forward-compatibility.

For @-patterns, this should ideally also bind as binding @ (sub: T) as opposed to (binding @ sub): T when writing e.g. let binding @ sub: T;. This might be observable via type checking of things like let ref binding @ A(_): T = scrutinee; but fortunately binding: &T holds here and not binding: T so like with or-patterns above we might be good. Still, before stabilizing F-bindings_after_at (e.g. binding @ Some(sub_binding)) we should endeavor to be more sure.

[chenyukang]

Closing now for #101728