Enable increasing complexity in rule-based scraper

[lorey]

Currently, the rule-based scraper tries all potential css selectors at once. It would be more performant if we increase the css selector complexity step by step though, so we first try single node rules like div.item and something like .menu > div.item.company if the simpler rules don't work.

[lorey]

This is still open, currently rules get generated with more complexity, but increasing complexity stops at selectors with two levels, e.g. .test > .value.target.box.