cpu time per mpi task scaled dow by the number of omp processes participating into it

test merge. I changed the print-out of the cpu-time printed by print_this_clock routine so that it does not shows the sum of the times of all the cpus contributing to an openMP thread but the average per cpu. it should make easier to compare timing of calculation made with different number of threads and should be always of the rder of the corresponding wall time. It might not work properly if the machine use hyper-threading.