* KRunner feels snappier overall, especially when dealing with a result set that hardly changes, e.g. using the calculator runner
* Ran model test and fixed most of its complaints. However, when updating a category, we append the new items, replace the entire list, and afterwards signal a data change, which model test doesn't like. It checks for whether the data before the inserted area didn't change, which it did at this point but that's why we signal a data change afterwards :)
* Running results still works
* Replacing the query string when requested still works, e.g. clicking on a calculator runner result changes the query string to that number
* Dragging results from KRunner to elsewhere still works. I noticed dragging applications from krunner to kickoff doesn't work but didn't check if that is a regression or because KRunner closes when kickoff opens
* Invoking runner actions (e.g. "run in terminal") still work