Investiate improving moc's pp logic

moc's preprocessor logic mostly works, but there are still some issues. One finding from QTBUG-124288 is that Preprocessor::tokenize doesn't actually split on white-space, which looks rather wrong – most of the time, that's not an issue, because we just skip over the tokens, but in certain instances, it can cause issues (e.g. when using token pasting).

Another part requiring improvements would be better handling of what's allowed in string and number literals (as in, the prefixes and suffixes).