Saving facets produces MPI hang with np=4, 2_5D_Wind_Farm

[eyoung55]

When the domain is subdivided and/or refined (the exact cause isn't 100% clear) past a certain threshold, it seems that some processes have empty values associated with certain marked boundaries, leading to error when trying to save facets. This could also be related to losing parent-child facet relationships when refining in parallel. In any case, the starting point should be the 2_5D_Wind_Farm regression test, which shows this behavior when np = 4 but not when np = 2.