Kanahiro/cloud-optimized-geoparquet — reverse-engineered prompt

Build me a clear specification repo for a draft format called Cloud Optimized GeoParquet Profile.

I want it to explain, in plain language, how a normal GeoParquet 1.1 file can be arranged so maps can load quickly from cloud storage. The key idea is that row groups are ordered from coarse detail to fine detail, with metadata showing where each level ends, so a reader can fetch only the first chunks it needs using HTTP range requests or object storage.

Make sure the spec is careful that this does not simplify, duplicate, or change geometries. Each feature should appear exactly once and the file should still work as ordinary GeoParquet for readers that do not understand the extra metadata.

Include a README that explains the motivation, design influences like Cloud Optimized GeoTIFF and point clouds, the current draft status, links to the full spec, and a short roadmap for producer, reader, and benchmarking work.