rewritten MultiRunSim to allow for minimum of rsh calls
MultiRunSim beforehand went through each fragment, one after the other, and called mpirun to commit it to a node. As the calculations with MPQC are so fast (roughly a second), this lead to a lot of rsh calls. The inetd of the job distributing node subsequently hanged itself after a short while. This was especially a problem in the BOSSANOVA scheme.
Now, we split the total sum of fragments up into as many packets as there are processor groups and commit them by single rsh call with all jobs in the packet concatenated with ";". Hence, we have a lot less rsh calls and now hangup of inetd. Note however, that for small molecules this still may lead to above described behaviour, i.e. if per packet there remains only one or two jobs. This cannot be overcome by any other mean than to lower the number of processor groups.
