Add run dependency on cuda-nvcc when using CUDA 12

[traversaro]

jax has a runtime dependency on cuda-nvcc. Before CUDA 12, there was no cuda-nvcc package in conda-forge, so users had to install it in a different way (see https://github.com/google/jax/blob/jaxlib-v0.4.25/docs/installation.md#conda-installation for the jax docs on this, that suggest to install cuda-nvcc via the nvidia channel.

Since we now also have CUDA 12, the cuda-nvcc is available in conda-forge, so in the CUDA 12 case, we can just add a run dependency on it (see https://github.com/google/jax/blob/jaxlib-v0.4.25/setup.py#L132 for a similar dependency in the upstream package).

The existing build was also missing a run dependency on cuda-cupti, this will also be solved by this PR as we will build against the cuda-cupti-dev package that now has the correct run_exports (see https://github.com/conda-forge/cuda-cupti-feedstock/issues/9 and https://github.com/conda-forge/cuda-cupti-feedstock/pull/10).

Fix https://github.com/conda-forge/jaxlib-feedstock/issues/239 .

Checklist

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

[conda-forge-webservices[bot]]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[traversaro]

@conda-forge-admin, please rerender

[traversaro]

fyi @diegoferigo @flferretti @giotherobot @lorycontixd

[xhochy]

@isuruf @jaimergp Do you know what the issue with cirun.io is? Shouldn't it be OK with the global .cirun as a config.

[isuruf]

Should work now. Thanks @aktech for fixing this

[ngam]

What triggered osx-arm testing all of the sudden?

@conda-forge-admin, please rerender

[ngam]

@isuruf @xhochy I don't know if the merge should come from either of you to ensure the Cirus thing works. I think we should merge this. We will deal with the osx-arm stuff in another PR.

[xhochy]

Yes, let's deal with osx-* separately.

[traversaro]

Thanks!

[traversaro]

The existing build was also missing a run dependency on cuda-cupti, this will also be solved by this PR as we will build against the cuda-cupti-dev package that now has the correct run_exports (see conda-forge/cuda-cupti-feedstock#9 and conda-forge/cuda-cupti-feedstock#10).

This turned out not to be true, as the fix fixed 12.4. builds, but all builds are using 12.0. due to cuda pinned to that version. See https://github.com/conda-forge/cuda-cupti-feedstock/issues/14 .