global parameter to control number of threads
- Introduce global parameter NCPUmax to specify maximal total number of CPUs available for fork(). Introduce interfaces to get this value as xf_cpu(int NCPU). Perform fork and update NCPUmax for the child processes by xf_fork(int NCPU).
- All calls to fork() are replaced by xf_fork(NCPU); ncpu/nthreads is updated to use xf_ncpu()
Tests are not fully completed (seem to work fine for profiler + RT combination)