Add support for punctuation in generated textmate grammars [#3337]

[iTob191]

I've implemented the change as outlined in #3337 (comment).

[szarnekow]

It seems as if we could use the same implementation for bother regex rules. The only meaningful difference seems to be the s.matches("\\w+") vs its negation and the name of the match rule.
Regular keywords should still work if we funnel them through escapeAndAddWordBoundaries

Would that work?

[iTob191]

That works and I've updated the PR accordingly. Only downside is that the generated regex for the keyword.control rule is less optimized: (\\ba\\b|\\bb\\b|\\bc\\b) instead of \\b(a|b|c)\\b

[iTob191]

After extending my grammar with Xbase support which brings in a lot more punctuation characters, I've noticed that this PR breaks highlighting of comments. The problem is that Xbase introduces / as a keyword which is then picked up by the punctuation rule and prevents other rules further down (like the comment rules) from being chosen. To resolve this, I have pushed an updated version in which the punctuation rule is moved to the bottom.