Make sure the alma images have different names

[isuruf]

Checklist

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

[jakirkham]

The original idea here was to use a multiarch image. It's possible there's some else still need to make that work well. Am happy to take a look if needed

[isuruf]

It doesn't work at the moment. We only use the docker image name in conda-smithy and just saying linux-anvil-alma would bring in arm64 image on a x86_64 runner. Also when emulating the full system, we need to be able to say we need the ppc64le docker image on the x86_64 runner which is not possible right now as well.

[xhochy]

This could be done by also passing in the --platform switch on docker run to select the underlying architecture but this didn't work well for me in the past. It often fell back to the native architecture when I had selected a non-native/emulated one. Using the sha256 digest works reliably to select non-native platforms but I don't see a way to transfer that easily into our build scripts.

Thus I'm also in favour of this PR as it keeps the images easily distinguishable and works very reliable in selecting the execution architecture.

[isuruf]

Ping on this @jakirkham

[isuruf]

Ping