Add type annotation syntax for lambdas

[j-hui]

Right now, if we want to type-annotate a lambda, e.g., fun x { x }, we need to write it postfix, i.e.,

(fun x
  x): a -> a

However, this can get kind of cumbersome. @sedwards-lab suggested that to keep the syntax uniform, we should also support the following:

fun (x: a) -> a
  x
fun x : a -> a
  x

This should of course generalize to lambdas with multiple arguments.

[sedwards-lab]

Yes. And it should be done using the same machinery somehow. let f x y z = ... and let f = fun x y z -> ... should be treated identically. Should we do the desugaring before understanding the type annotations?

[j-hui]

They are separate in the AST, but are desugared while lowering to IR.

What do you mean by "understanding" the type annotations?