C++: Exclude custom vprintf implementations from primitiveVariadicFormatter.

[geoffw0]

This is my long awaited solution to the problems we've been seeing with cpp/wrong-type-format-argument producing false positives in the presence of certain printf implementations (see https://lgtm.com/projects/g/microsoft/ChakraCore/alerts/?mode=tree&ruleFocus=2160310550).

As it happens this change removes ALL results on ChakraCore for cpp/wrong-type-format-argument (and also cpp/non-portable-printf, which is not on LGTM anyway). I believe that is what we want. Results are preserved on the other projects I tried (BOINC_boinc and godotengine_godot) so I believe the change is functioning as intended.

I went with exists(f.getBlock()) rather than exists(f.getFile().getRelativePath()) because, in my tests, the former would exclude things I don't think we really need it to (where the header has been copied into the source directory but the corresponding source has not?).

[jbj]

The tests failed.

[geoffw0]

Explanation of the test changes:

• the swprintf function in the test was no longer considered a FormattingFunction, because it depends on vswprintf, a vprintf variant with an implementation in the snapshot.
• some further results were affected because getAFormatterWideType no longer recognized that the snapshot uses char16_t as a wide type at all, and falls back to the default wchar_t (this sort of thing is less likely to change in a bigger / real world snapshot).

In my opinion this effect devalues the test, which is about identifying when wide strings are char16_t, so I've fixed it by changing the test. I'm not entirely sure why the function had an implementation anyway. Happy to add further testing if anyone feels we are now missing something.