Revert "[CSDiagnostics] Don't mention special names in requirement diagnostics"

[xedin]

Reverts #19447

[xedin]

@swift-ci please smoke test

[xedin]

@jrose-apple Maybe this is not actually worth reverting because note is going to point to declaration in question, with one caveat - if it's implicit initializer it would point to the type where requirement was declared...

[jrose-apple]

In plenty of contexts notes might not be displayed without clicking into an error, and in the case of Xcode they're often not displayed at all today. :-( The main error should be informative without having to do that.

If you want to clean up the obviously useless cases, you could say "if there's a special name and zero or one unlabeled parameters" as your condition or something.

[xedin]

Sure, that makes sense, I'll change the condition to only print names if they have at least one label instead of reverting.

[jrose-apple]

Ah, that's not quite the condition I was saying. If you have two unlabeled parameters, it would be good to know that you have an unusual subscript / non-converting initializer. (Even zero vs. one parameters for an init might still be interesting.)

[xedin]

Yeah, I think it might only make sense to really filter out special + no parameters because it would just be initializer 'init'

[xedin]

Talked to Jordan offline, and it seems like better option here would be to enable empty label printing for special names, which would make initializer 'init' be printed as initializer 'init(_:)', so I'm going to revert my changes and do it another way...