Reported CPU times are incorrect when used with multi-threading

[dweindl]

Currently, the value of Solver::getMaxTime() is checked against std::chrono::system_clock, which measures wall time. This is probably not what we want and can cause issues if a program is paused or the cpu is oversubscribed. Should be changed to std::clock. The other time measurements reported in ReturnData are already using clock.

[dweindl]

Just using clock() won't do the job for parallel simulations. Would need separate Windows- and POSIX-code to get thread-specific CPU time. boost::chrono::thread_clock would work, but we don't want to depend on boost. Probably not worth the trouble at this point.

[dweindl]

Well, that also means that all currently reported cpu times obtained from from runAmiciSimulations(..., nthreads=N), N>1 are probably garbage. So we should do something about it...

My current preference would be to use boost if boost is available, and to add a warning that in other cases, those numbers are most likely incorrect (or just set them to nan then?).