add benchmark for background parser

it's pretty slow, 1000 documents takes ~2s even with our dummy language
thats used in the unit tests

one thing thats pretty slow is the iteration over the url sets in m_documentsForPriority.
I've toyed around with it but even replacing it with a QVector wasn't faster (and
made the code much more complicated due to manual set-logics...)

has anyone some ideas on how we could improve this situation?