Use new coverage tasks

Use all .set-files necessary to include all tasks of the sv-benchmarks that have a coverage property declared.

This, at the moment, leads to multiple tasks being included twice, and a warning is shown. I think this may lead to wrong benchmark data?

If it does, we'll have to create our own .set files or ask benchexec to handle multiple task definitions as a union (instead of whatever it is doing now).
We can't solve this by declaring different run definitions for the different categories, because this may still execute some tasks multiple times and thus multiply the score on these tasks.