redo 'without' operator parsing so it's within parseType to match C++ parser

[kavon]

I decided to change the parsing on the C++ side to instead look for suppression in parseType as it yielded a better implementation. As such, "suppressed" inherited entries are now gone and you can now have general suppressed types being parsed.

[kavon]

@swift-ci please test

[kavon]

@swift-ci please test

[ahoppen]

I am not following the evolution proposal but do we really want to support ~ in any place that supports types. As it stands right now, once we integrate the new parser into the compiler, ASTGen will need to diagnose that ~ cannot be used for every syntactic position that supports types.

[kavon]

Eventually it will need to be supported in where clauses and in parsing generic parameters like <T: ~Copyable>. At the moment TypeCheckType.cpp flags those appearances as illegal, so I think ASTGen can safely just map it over to a SuppressedTyprRepr as in swiftlang/swift#64699

[kavon]

So you can think of this as an improvement in terms of diagnostics, since now we give a good diagnostic complaining about uses of ~Type in other places, as people are bound to try it elsewhere. Without this change they'd get really bizarre parsing errors. Those diagnostics happen during type checking, though, not parsing. See test/Sema/without_implicit_conformance.swift in that PR for examples of the messages.

[kavon]

@swift-ci please test windows