feat: if.scikit-build-version

[henryiii]

This adds an if.scikit-build-version, and makes sure it can be used to gate features not yet implemented. This should help with #769 by providing a way to support old scikit-build-core's on Python 3.7 even after we drop support.

Also fixes auto minimum-version to respect markers.

[LecrisUT]

The plain version naming is so ambiguous though. What about skbuild-version or skbuild-core-version to mirror the CMake variable?

[codecov]

Codecov Report

Attention: Patch coverage is 92.30769% with 2 lines in your changes missing coverage. Please review.

Project coverage is 83.87%. Comparing base (0a8a352) to head (54ff64a).
Report is 52 commits behind head on main.

Files with missing lines	Patch %	Lines
...rc/scikit_build_core/settings/skbuild_overrides.py	91.66%	2 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #851   +/-   ##
=======================================
  Coverage   83.87%   83.87%           
=======================================
  Files          74       74           
  Lines        4347     4361   +14     
=======================================
+ Hits         3646     3658   +12     
- Misses        701      703    +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[henryiii]

I was trying to match minimum-version, which also doesn't have any extra qualifier. Other versions, like cmake.version or python-version have qualifiers. That was the thinking, anyway. It's also rather important, so scikit-build-core-version or scikit-build-version seems really long, especially for something this integral. I've avoided "skbuild" in pyproject.toml.

Admittedly, minimum-version is tool.scikit-build.minimum-version, while this is tool.scikit-build.overrides[].if.version, which is a bit more disconnected from the "scikit-build" part.

[henryiii]

Would if.__version__ be clearer? Or if.core-version? We've mostly used "scikit-build" in the settings, but I could see confusion here if scikit-build-version was used.

[henryiii]

scikit-build-version isn't terrible either, I suppose.

[LecrisUT]

while this is tool.scikit-build.overrides[].if.version, which is a bit more disconnected from the "scikit-build" part.

☝️

Length wise I think it's fine as it will always be in its own block with only if. or if.*. prefixed.

scikit-build-version isn't terrible either, I suppose.

👍. Not much longer than python or minimum