(~) and (!) are parsed inconsistently in types (plus documentation warts)

(Spawned from #10056 (closed)?replyto=41##15457 (closed))

~ and ! are slightly special in the parser to allow strict annotations and lazy annotations (with -XStrict) in data types. As a result, we initially parse all uses of ~/! as bangs, and we use a post-parsing pass (mergeOps/splitTilde) to figure out which uses of ~ are actually meant to refer to the equality type constructor.

There's a couple of unsatisfying things here:

1. splitTilde handles ~, but not !. This means that any use of ! as a type operator will not work, as evidenced by this GHCi session:

λ> type a ! b = Either a b

<interactive>:1:6: error:
    Malformed head of type or class declaration: a !b

We should update splitTilde to handle ! as well. (And perhaps give that function a new name to reflect its more ambitious goals.)

1. Note [Parsing ~] is supposed to explain all of this hullabaloo, but it does a rather poor job of it. Let's add some more prose to it.

Trac metadata

Trac field	Value
Version	8.4.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler (Parser)
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture