Use faster ccache, MR to feature/add_xenial

Hi @gladk, there were some problems with docker on 4pak. The ccache approach which you used did not work, and effectively every compilation was without ccache. 4pak has 48 cores, and can accept max 5 jobs. I've set this limit high so that tests & checks are run faster. While compilation with ccache barely uses CPU power. But in this case we had 3 compilations running in parallel on 4pak, each using 43 CPUs. Docker was so slow that it stopped responding, and even failed to kill those jobs after 2hour limit has passed. I had to restart docker & gitlab-runner manually on 4pak.

So I thought, better to make sure that ccache is used. In this MR I enabled it. I tested this first on xenial only and saw that no CPUs were used during compilation. We can let this MR sit here and test to make sure. But I think this should work.