add test for correctly resolving stdlib

[h-vetinari]

Closes #1840, needs conda-build 24.3; tested locally with artefacts from conda-forge/conda-build-feedstock#220

[h-vetinari]

I'm actually surprised that this passes with conda-build 24.1.2; I thought that conda/conda-build#5195 would have been necessary, but I guess by adding c_stdlib{,_version} to always_keep_keys, we're not actually dependent on that PR after all... CC @mbargull

[jakirkham]

This is great! Thanks Axel 🙏

[mbargull]

I'm actually surprised that this passes with conda-build 24.1.2; I thought that conda/conda-build#5195 would have been necessary, but I guess by adding c_stdlib{,_version} to always_keep_keys, we're not actually dependent on that PR after all... CC @mbargull

Well, sure, without conda/conda-build#5195, the key is not recognized as used; if you override the logic with the (conda-smithy-specific) always_keep_keys thing, you don't depend on it being recognized, of course.

[h-vetinari]

Yeah, I didn't know the inner workings of conda-{build,smithy} well enough to understand that, sorry...

[mbargull]

Ah, no need to apologize!
We needed the fix in conda-build anyway since we have support uses outside of conda-smithy rerender workflows ;).