Pip 22.1.0 breaks many recipes

[henryiii]

Pip added unavoidable build requirement verification when building in no build isolation mode in pypa/pip#10886. This breaks many packages on conda-forge, namely almost any that depend on numpy (see pypa/pip#11116 and pypa/pip#11118), and any that depend on cmake or ninja, which do not install Python packages.

This was almost immediately reverted and instead made an opt-in check. But there hasn't been a new Pip release yet, so I think we should mark pip 22.1.0 as broken (or apply the patch in pypa/pip#11117). @conda-forge/core?

Some examples of the error message:

pip._internal.exceptions.InstallationError: Some build dependencies for file://$SRC_DIR are missing: 'cmake>=3.12', 'ninja'.
ERROR: Some build dependencies for file:///home/conda/feedstock_root/build_artifacts/dwave-preprocessing_1652872033983/work conflict with the backend dependencies: numpy==1.22.3 is incompatible with numpy~=1.21.0.
ERROR: Some build dependencies for file:///Users/rgommers/code/tmp/scikit-learn are missing: 'oldest-supported-numpy'.

A example feedstock that ran into this: conda-forge/resample-feedstock#1 (comment)

[beckermr]

I'm ok with either solution above.

[jakirkham]

Also ok with either option and happy to review.

The pip developers also indicated they planned to release soon ( pypa/pip#11117 (comment) ). So held off doing anything assuming they understood this is urgent

[ocefpaf]

Marking them as broken in a PR in a moment.

[ocefpaf]

Done in conda-forge/admin-requests#439