uv: Deduplicate all files in the wheel cache

The uv project is extending content-addressed caching from whole wheels to individual files, deduplicating package contents in the wheel cache by storing files under BLAKE3 hashes and hardlinking them back into wheel archives. For CIOs and technology leaders, the business value is lower infrastructure footprint and better cache efficiency—about 10% less local cache usage in testing—while preserving install behavior and keeping warm-install performance effectively unchanged, with only a small cold-install slowdown of under 4%. This signals a strategic move toward more storage-efficient dependency management that can reduce operational overhead at scale, especially in environments with large Python estates and repeated package installs.