Cleanup

[dshean]

Testing with large geojson (~18000 km2), which involved 12278 tiles (2x2 km):

• Greatly speeds up the EPT reader preparation for each tile. Still serial, but eliminates the need to loop through 2201 EPT polygons for each tile. Limits initial read of EPT polygons to the user-defined AOI polygon. Then does intersection once, and only adds valid EPT files to the reader.
• Fix bug that only allowed 9999 pipeline json files to be created
• Clean up import statements
• Clean up helpstrings, description and comments
• Clean up print statements

Several related issues require further attention, and some additional refactoring.

[dshean]

This was a branch from the osx-64_support branch. It could be merged directly onto main.
We can close and open PR directly on main.

[dshean]

Ah, the existing tiling approach uses bbox, but doesn't check for intersection with the original AOI. Produces A LOT of unnecessary tiles. See the first two tiles (southwest) here (red is AOI, green is EPT index).

Will push a fix for this. We should consider creating geodataframe of tile extent polygons up front, rather than iteratively creating.

[dshean]

We should also consider preserving the tile row/col in the pipeline filenames.

[dshean]

This reduces the number of readers to process from ~11600 to ~6900.

Unfortunately, it means that the tile numbers are inconsistent with the pipeline numbers. Not a showstopper, but not ideal.

I think a better approach here is to create the tile polygon index for the entire bbox, intersect with the user AOI, then store all of relevant metadata in the gdf for the tile, instead of creating and maintaining 4 additional lists for pointcloud_input_crs, readers, extents and original extents. Just return the GDF and expand elsewhere, instead of the lists. Keep the same tile ID.