clean up ccm direcotry between run

[fruch]

otherwise next run would try to reuse existing clusters and they are already stopped and killed