Spike: generate and shelve PMTiles data

[thatbudakguy]

The open-source PMTiles format for GIS data is designed to be served directly from cloud storage or a CDN, to remove the need for GeoServer for content delivery. Notably, it is the first format designed to handle both raster and vector data; it is a generic "tiles of anything" binary that supports HTTP range requests. For all types of content we ingest (and many that we don't), there is a path to conversion into a PMTiles archive.

While there are no Stanford-local applications that consume this data yet, there are examples of service providers in the wild (notably Felt) who have used this approach to radically streamline their GIS data processing pipeline, since you effectively only need to produce a single format regardless of the input data type.

In this future, we could decommission the GeoServer infrastructure entirely and serve all GIS data directly from stacks. While that is probably a long ways off, we can hasten its arrival by starting to generate these files and make them available to downstream applications.