Plotting Boundaries in train_model.py --eval not just interior

[sadamov]

With milestone v0.4.0 we are releasing realistic boundaries. For that purpose the plotting functionalities need to be updated. Currently they only plot the interior domain, with a semi-transparent boundary around it. In the new approach the boundary and interior data comes from two separate datastores.

Some considerations:

• Using domain_cropping in the config will keep only the points at the given distance from the convex hull of the referenced interior data. This set of points will in general not lay on a regular grid.
• There is no boundary mask in neural-lam right now. All points in the boundary datastore are considered boundary input points.
• As plots anyhow are in the crs of the interior, boundary points would still not lay on a regular grid (if they were originally in e.g. a regular lat-lon grid).

I started working on this but got lost in the x,y,lat,lon,projections,coordinates. When are they 1D, when are they 2D, stacked unstacked. Anyways :D here is my working branch: https://github.com/sadamov/neural-lam/tree/plot_boundaries But unfortunately I don't have the bandwidth right now to continue working on it. A typical example plot looks like this: