Fix grounding line mask definition when creating landice meshes

[trhille]

This merge fixes grounding line mask definition, which previously marked the entire ocean as the grounding line used for setting cell widths. It also saves maps of the margin and grounding-line masks, which will be helpful for debugging purposes when new maps are added or mask definitions are changed.

Checklist

• Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

[trhille]

Testing

I tested this on the Greenland mesh with 4–40km resolution. Using thee previous definition of the grounding-line mask, we get a mask that is really a union of open ocean and floating ice masks:

Using this update, we get:

These maps are now saved as pngs in the mesh_gen case directory.

Note that this doesn't make much practical difference for the Greenland mesh, since we take the minimum of the distances to the grounding line and ice margin. The cell area using the previous grounding line mask is shown on the left (nCells = 29953), with the changes in this merge shown on the right (nCells = 29748 ):

[trhille]

I'm not sure this is the most straightforward way to define this. I have trouble thinking through the logic with these rolled arrays. If @matthewhoffman or @xylar has ideas about making this simpler, I'd be happy to hear them.

[matthewhoffman]

I also find these things confusing. ChatGPT suggested this as an alternative:

shifted_grounded = np.roll(grounded_mask, n, axis=[0, 1])
shifted_float = np.roll(float_mask, n, axis=[0, 1])

not_grounded_mask = (~shifted_grounded) & shifted_float

Side question: would it be more accurate to rename not_grounded_mask to neighbor_not_grounded_mask?

[matthewhoffman]

@trhille , the maps look great and I like that you are now saving them out. I am happy to approve & merge once we're happy with that question about the logic for the masks.