Use griddata's convex hull to determine the valid interpolation range

[vincentpost]

This part can probably be skipped:

data/src/nhflodata/data/mockup/bodemlagen_pwn_2024/v2.0.0/botm/interpolation_helper_functions.py

Lines 866 to 878 in 882a24d

	within = gpd.sjoin(gdf_result, boundary_gdf, predicate="intersects")
	mask_inside = gdf_result.index.isin(within.index)
	# Nearest-neighbor fallback for NaN points inside the boundary
	mask_nan_inside = mask_inside & np.isnan(interpolated)
	if mask_nan_inside.any():
	nearest = griddata(src_xy, src_values, all_xy[mask_nan_inside], method="nearest")
	interpolated[mask_nan_inside] = nearest
	gdf_result["value"] = interpolated
	# Clip: set values outside boundary to NaN
	gdf_result.loc[~mask_inside, "value"] = np.nan

reasons: griddata automatically creates a convex hull when using 'linear' and clipping to boundary_gdf is not strictly necessary. Within convex hull interpolation result will be used, outside of it values from REGIS will be used. Lines will be removed after testing if the interpolation result is indeed as envisaged.

[bdestombe]

Note that in this case, the bottoms of the aquitards require the top-aquitard and thickness-aquitard data to both reside in the convex hull. I recon this is correct behavior.

[bdestombe]

See #60. BOTM locations where only thickness of aquitards were available now also enter botm.geojson