MPAS should generate graphs inline

[jedwards4b]

I don't think that there is a good reason MPAS ATM cannot generate it's decomposition
graph at runtime. This would prevent issues like ESMCI/cime#3092

[mgduda]

I don't disagree that it's technically possible to generate mesh partitions at runtime. ParMetis does have an API that we could use, though I haven't worked with it much at all. This would add an additional library dependency, and I think the work to implement this capability wouldn't be entirely trivial. In any case, I'll see whether this is something that we could do in the not-to-distant future.