Produce DEM layer from EPT using Spark

[echeipesh]

The driving use case behind DEMRasterSource development is to produce dynamic DEM raster from triangulation of the pointcloud subset returned from Entwine tree query. The intent is to use Entwine trees as authoritative data source and minimize the work required to maintain the derived DEM product.

Two problems currently exist:

• Generating triangulation on the fly is feasible but still relatively expensive
• PDAL Pipeline is not thread safe, limiting parallelism and throughput of such service

An alternative approach may be to have an AWS EMR or AWS Batch triggered after EPT tree creation and update to generate the DEM product and store it as STAC indexed COGs (for instance). The upkeep can be minimized due to automated nature of the job and for high query volumes the raster layer becomes a cache into triangulation.

DEMRasterSource can be used to prototype this approach: parallelize the generation of full DEM from highest resolution of EPT tree.

Note: this issue should be evaluated against existing tools to perform similar workflow. At the moment I'm not sure what they are, but a quick look should reveal.