Allow parallel processing options for multiple CFD simulations

[jforthofer]

When a user starts a CFD simulation with multiple runs (ie. more than one input speed and direction for the domain average run) we have several options for spreading the runs over the processors.

1. Do the runs in serial, but spread each serial run out over all the processors. This is the current strategy for 3.0.0. Advantage is it keeps the RAM as low as possible, disadvantage is it may not be fastest.
2. Let one run be attached to one processor. Advantage is it might be faster than number 1. Disadvantage is it takes more RAM and might run out of RAM.
3. Spread runs and processors somewhat evenly. In this method, multiple runs could be done at the same time, with each run potentially using multiple processors. For example, with 4 runs to do and 8 processors, we could do each run with 2 processors at the same time. We could also attempt to code in logic to try to do this without running out of RAM. This would be approximate, and could fail if the estimated RAM used by WindNinja is underestimated or if the user consumes more RAM with an outside process during running, etc. Advantages of this if it worked would be basically the best of both worlds, not run out of RAM and do the most concurrent runs possible on the computer.