Add critical land blockages to global_ocean tests

[xylar]

Adds support for transects that must be land to the new global ocean setup in COMPASS

[xylar]

These changes will only work along with:
MPAS-Dev/geometric_features#103
and
MPAS-Dev/MPAS-Tools#247

[xylar]

Testing

So far, I have only tested on a QU240 mesh and that's a bit of a hacky test because I don't have Jigsaw yet so I just copied the mesh for a global_ocean_V1 test case.

Here is the masking I see:

At this resolution, it's pretty hideous but the land mask has been thickened a bit around the Peninsula (unnecessarily in this case) to make sure there's no through-flow.

[xylar]

@mark-petersen, could you try this with the EC60to30 setup, since you're ready to do the full Jigsaw workflow? Then, we could try CONUS if that works.

[mark-petersen]

This is a mega-test, where I test multiple components at once for the EC60to30 mesh:
MPAS-Dev/geometric_features#49
MPAS-Dev/geometric_features#103
MPAS-Dev/MPAS-Tools#247
with this one. On badger, I used:

  46: -o ocean -c global_ocean -r EC60to30 -t init
cd 
![screen shot 2019-03-05 at 5 25 42 pm](https://user-images.githubusercontent.com/4036816/53846848-bd361700-3f6b-11e9-8525-f7b376aa3996.png)
/usr/projects/climate/mpeterse/repos/model/pr_test/testing_and_setup/compass
./setup_testcase.py -f general.config.ocean_ba_xylar_mesh_test --work_dir /lustre/scratch4/turquoise/mpeterse/runs/190305_test_xylar_mesh_culling -n 46

I ran all the way through init and part of spin-up. Everything went through.

On these images, we have left to right EC60to30 from

1. Mesh used in E3SM V1 campaign, named oEC60to30V3
2. Jigsaw-created mesh from current repo, before all these PRs.
3. Jigsaw-created, with all these PRs.

Colors show bottom depth.

These images show that the new features work, or at least don't cause harm. Note that Nares Strait is open on upper left. Thank you @xylar for your dedicated effort on this.

[mark-petersen]

It's also very nice to look at the mask files, as they show what cells were removed or forced in, and why. Here is the file

critical_passages_mask.nc
	int transectCellMasks(nCells, nTransects) ;

While this file shows the cells that are removed because otherwise sea ice can't flow through:

land_mask_final.nc
	int regionCellMasks(nCells, nRegions) ;

[mark-petersen]

If anyone wants to explore these files and make more plots, be my guest.

Grab this tar file:

/lustre/scratch4/turquoise/mpeterse/runs/190305_test_xylar_mesh_culling/ocean/global_ocean/EC60to30/init/tar_file/EC60to30_test_xylar_mesh_culling_190305.tar

or see the directory here:

/lustre/scratch4/turquoise/mpeterse/runs/190305_test_xylar_mesh_culling/ocean/global_ocean/EC60to30/init/initial_state

[mark-petersen]

The EC60to30 images were enough to convince me these four PRs are working together properly. If it turns out we need further refinement, it will go into another PR. Thanks again @xylar.

[xylar]

The screenshots in this PR should not be studied in too much detail, as the land mask is corrupted, see: #176