Fragments can be used only in another lexer rule error

[KvanTTT]

A more clear error message to avoid confusion if fragment lexer rules are used in parser rules; fix #3016, fix #2886

[parrt]

hi. This is a bit hard to evaluate since there seem to be a lot of changes but really they are just comment changes. Can you undo all of the unnecessary changes so I can look at the difference? BTW, I'm pretty nervous about introducing a new imaginary token.

[KvanTTT]

Can you undo all of the unnecessary changes so I can look at the difference?

Ok.

BTW, I'm pretty nervous about introducing a new imaginary token.

It's possible to implement without introducing a new imaginary token type but such error will work only with combined grammar, i.e. the test testFragmentRuleInParserRuleInSeparatedGrammar will not work.

The current implementation adds fragment tokens to .tokens file and analyzes them in the parser generation stage if grammars are separated:

TOKEN=1
FRAGMENT=-3
FRAGMENT2=-3

[parrt]

Hmm... in the end I don't have the motivation to think carefully about these things and any kind of changes that affect end-users should be done after a lot of thought. I think we should focus on the targets.