Optimization of some rules and memory reduction

The first commit reduces the number of members in the rule class

./bin/highlighter_benchmark:
Before: 26761ms - 77200K
After: 26247ms - 73812K

The third commit improves DynamicRegExpr to minimize the use of QRegularExpression (and sometimes not to use it at all).

./bin/highlighter_benchmark:
Final: 22432ms - 73064K

I noticed earlier that we were in C++14, so I could improve the readability of dataRuleVisit() a bit by using lambdas with auto parameters.