Lint against unusual constraint orderings #2343
Description
statsmodels used to have a constraint scipy >=1.8,!=1.9.2, which got flipped to scipy !=1.9.2,>=1.8 in conda-forge/statsmodels-feedstock@12ba4cb (by the bot! 😱).
This order prevents the already complicated _pin_stricter in our repodata patches from working correctly, see conda-forge/conda-forge-repodata-patches-feedstock#1044.
IMO it would be a better trade-off to just prohibit this kind of pattern through the linter, rather than make our repodata patching code even more complicated.