Consider using the pyogrio library to speed up vector reading & writing

[robyngit]

Today I learned about the pyogrio package, created by the geopandas developers. It enables much faster reading (5-10x faster) & writing (5-20x faster) of vector formats than with geopandas alone. Using pyogrio.read_dataframe and pyogrio.write_dataframe would be super advantageous in all three of our packages: viz-staging, viz-raster, and viz-3dtiles. The docs also outline additional features that would be useful in speeding up our workflow:

• Filter records by spatial extent: In viz-staging, we could select and read only polygons that intersect with the tile bounding box, rather than reading in the entire file at once.
• Read basic information about a data layer - get CRS, features, num polygons, etc. without reading in the entire file. This would allow us to convert the CRS of the tile bounding box to the CRS of the data to enable the above
• Filter records by attribute value - For viz-3d tiles, we could read in only polygons where centroid_within_tile is true.
• Reading from compressed files / archives - we could add support for reading in .zip files without the need for unzipping them

However, note the warning:

Pyogrio is still at an early version and the API is subject to substantial change.

[robyngit]

Quick note that as of geopandas 0.11, you can set the reading & writing "engine" to pyogrio instead of the default/slower fiona. This is an easy step we could implement to make the workflow a little faster, but it would need some testing.