Issue handling microach variants in recipes

[matthewfeickert]

trafficstars

I'm not sure how to describe this super well, but in https://github.com/conda-forge/awkward-cpp-feedstock/pull/67 we noticed that the conda-forge auto-tick bot was not updating the awkward-cpp feedstock.

The state of the feedstock's recipe at that point is here: https://github.com/conda-forge/awkward-cpp-feedstock/blob/174204689d45a9e491b2bd845d15e7cf5e30697b/recipe/recipe.yaml

In https://github.com/conda-forge/awkward-cpp-feedstock/pull/67#issuecomment-3073170539, @jaimergp correctly identified that the x86_64-microarch-level was not being handled correctly and suggested an additional guard, which was implemented in https://github.com/conda-forge/awkward-cpp-feedstock/pull/68.

https://github.com/conda-forge/awkward-cpp-feedstock/blob/448a31e93ef73aaf92c09f9993943ac90a845f8e/recipe/recipe.yaml

This then allowed for https://github.com/conda-forge/awkward-cpp-feedstock/issues/69 to correctly trigger an version update in https://github.com/conda-forge/awkward-cpp-feedstock/pull/70.

@jaimergp thinks this points to a bug in rattler-build, though I don't understand enough to be able to speak intelligently on this. I'm happy to provide more info here though.

cc @henryiii