Clear Python installation scratch space with uv cache clean

[woodruffw]

Summary

Fixes #20675.

There are two distinct parts here:

• uv python itself now has a provision to attempt to clear the scratch directory
• uv cache clean now tries to clear the scratch directory by default

I'm not 100% sure whether either/both is best -- I think clearing as part of uv python is arguably superfluous (IIUC there will functionally never be anything to clear, since the only way clear_scratch will be reachable is if an installation/download wasn't interrupted). Arguably therefore we should maybe only do this as part of uv cache clean.

Curious if @geofft has strong opinions on that as well 🙂

Test Plan

Added tests for various scenarios with uv cache clean + different states for the Python installation scratch directory.

[astral-sh-bot]

uv test inventory changes

This PR changes the tests when compared with the main base revision.

• Added tests: 7
• Removed tests: 0
• Changed suites: 2

uv::build: +5 / -0

Added:

• uv::build::cache_clean::clean_all_does_not_report_empty_python_temporary_directories
• uv::build::cache_clean::clean_all_python_temporary_directories
• uv::build::cache_clean::clean_package_preserves_python_temporary_directories
• uv::build::cache_clean::clean_python_temporary_directories_waits_for_installation_lock
• uv::build::cache_clean::clean_python_temporary_directories_without_cache

Removed: none

uv::python: +2 / -0

Added:

• uv::python::python_install::python_install_automatic_cleans_stale_temporary_directories
• uv::python::python_install::python_install_cleans_stale_temporary_directories

Removed: none

[woodruffw]

#20752 addresses the main root cause of this. We could still do the things here, but IMO only the uv cache clean half is probably worthwhile. Closing for now until we have more evidence that interrupts (rather than network errors) are the root cause.