Open-mp threading

[jhamman]

The current VIC MPI approach does not utilize threading for shared-memory on-node calculations. This might be expected to cause performance slow downs coming from a large number of MPI calls. This PR is introduces openmp threading to the vic_image_run loop over local_domain.ncells_active as proof of concept.

[bartnijssen]

@jhamman: I don't want to try this for the time being. This should be shelved till after we have a working VIC 5 release. I am closing the PR for now, because I think this is a bit of a distraction, especially since at this time we don not even know whether it will provide a speed up or not.

[jhamman]

That's fine @bartnijssen, I ran across this while working on something else and wanted to jot it down before I moved on and forgot how easy it could be. I labeled it someday so as to indicate its low priority. I'll open a separate issue (#522) but will not work on this again.