Fix linter error from 2.1.0 (due to `noarch: python`)

jakirkham commented 4 months ago

The linter does not like having noarch: python at the top-level with selectors ( https://github.com/conda-forge/xgboost-feedstock/pull/167#issuecomment-2251573632 ) ( https://github.com/conda-forge/conda-smithy/issues/1887#issuecomment-2244323850 ). Rewriting the recipe without selectors is a substantial lift and not worth it given the move towards rattler-build.

However if we don't have noarch or a skip at the top-level, a full matrix of Python versions is built out for all Python XGBoost packages (despite these packages being redundant). Example below taken from this CI run (attached log for posterity):

Packages generated by Python matrix in past CI run:

```json { "_py-xgboost-mutex-2.0-cpu_0.conda": { "recipe": { "channel_targets": "conda-forge main", "target_platform": "linux-64" } }, "_r-xgboost-mutex-2.0-cpu_0.conda": { "recipe": { "channel_targets": "conda-forge main", "target_platform": "linux-64" } }, "libxgboost-2.0.3-cpu_hce603c2_3.conda": { "recipe": { "__glibc": "__glibc >=2.17,<3.0.a0", "c_compiler": "gcc", "c_compiler_version": "12", "channel_targets": "conda-forge main", "cuda_compiler": "None", "cuda_compiler_version": "None", "cxx_compiler": "gxx", "cxx_compiler_version": "12", "target_platform": "linux-64" } }, "py-xgboost-2.0.3-cpu_pyh0a621ce_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.11.* *_cpython", "target_platform": "linux-64" } }, "py-xgboost-2.0.3-cpu_pyh78d450f_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.10.* *_cpython", "target_platform": "linux-64" } }, "py-xgboost-2.0.3-cpu_pyh995e691_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.8.* *_cpython", "target_platform": "linux-64" } }, "py-xgboost-2.0.3-cpu_pyhbc962bd_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.12.* *_cpython", "target_platform": "linux-64" } }, "py-xgboost-2.0.3-cpu_pyhc68e8b7_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.9.* *_cpython", "target_platform": "linux-64" } }, "py-xgboost-cpu-2.0.3-pyh099c497_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.12.* *_cpython" } }, "py-xgboost-cpu-2.0.3-pyh4f80c01_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.10.* *_cpython" } }, "py-xgboost-cpu-2.0.3-pyhac85b48_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.9.* *_cpython" } }, "py-xgboost-cpu-2.0.3-pyhb06c54e_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.11.* *_cpython" } }, "py-xgboost-cpu-2.0.3-pyhb8f9a19_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.8.* *_cpython" } }, "r-xgboost-2.0.3-cpu_r42h566b21a_3.conda": { "recipe": { "__glibc": "__glibc >=2.17,<3.0.a0", "c_compiler": "gcc", "c_compiler_version": "12", "channel_targets": "conda-forge main", "cuda_compiler": "None", "cxx_compiler": "gxx", "cxx_compiler_version": "12", "target_platform": "linux-64" } }, "r-xgboost-2.0.3-cpu_r43h566b21a_3.conda": { "recipe": { "__glibc": "__glibc >=2.17,<3.0.a0", "c_compiler": "gcc", "c_compiler_version": "12", "channel_targets": "conda-forge main", "cuda_compiler": "None", "cxx_compiler": "gxx", "cxx_compiler_version": "12", "target_platform": "linux-64" } }, "r-xgboost-cpu-2.0.3-r42hf5c67e3_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "target_platform": "linux-64" } }, "r-xgboost-cpu-2.0.3-r43hf5c67e3_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "target_platform": "linux-64" } }, "xgboost-2.0.3-cpu_pyh099c497_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.12.* *_cpython" } }, "xgboost-2.0.3-cpu_pyh4f80c01_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.10.* *_cpython" } }, "xgboost-2.0.3-cpu_pyhac85b48_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.9.* *_cpython" } }, "xgboost-2.0.3-cpu_pyhb06c54e_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.11.* *_cpython" } }, "xgboost-2.0.3-cpu_pyhb8f9a19_3.conda": { "recipe": { "channel_targets": "conda-forge main", "cuda_compiler": "None", "python": "3.8.* *_cpython" } } } ```

To clean this up, we skip any Python version except the minimum one that XGBoost supports. This ensures we build for exactly one Python. To get the syntax right we need to use Jinja.

Unfortunately that means we need to consolidate our other skips into this one, which includes the Windows CUDA 11.8 skip. So we do so and try to make the syntax as pleasing as we can within the constraints.

That all being said, this does keep the linter happy and it re-renders without generated unneeded Python builds.

Checklist

[x] Used a personal fork of the feedstock to propose changes
[x] Bumped the build number (if the version is unchanged)
[ ] Reset the build number to 0 (if the version changed)
[x] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
[x] Ensured the license file is being packaged.