Shade missing values when rasterizing Points

I am one of the developers of the [UXarray](https://github.com/UXARRAY/uxarray) package and we use Datashader and Holoviews for visualizing unstructured grids. The majority of our data is from climate model outputs, meaning that the data we are visualizing is mapped to the surface of a sphere and is projected for 2D visualization.

We support both Polygon and Point rasters, and one issue with the Point rasterization is that regions with lower point densities (such as the north and south poles when projected to 2D), lead to there being missing values near the poles. This can be seen when comparing the Polygon and Point rasters (look at the top and bottom)



While this is acceptable at high resolutions (the above grid is about 84 million points), the issue becomes much more noticable at lower resolutions

I'm wondering whether it would be possible to shade the missing values using some form of interpolation, such as nearest neighbor.

Start by reproducing the reported comparison between the Polygon and Point rasters at lower resolutions, focusing on missing values near the projected poles. Read the existing Point rasterization path and determine how nearest-neighbor or another interpolation should define those gaps; done means the missing regions are shaded consistently without regressing the existing raster outputs.

Written by the indexing model from the issue text.