Build and test with CUDA 13.0.0

[jameslamb]

Contributes to rapidsai/build-planning#208

• uses CUDA 13.0.0 to build and test
• moves some dependency pins:
  • cupy: >=13.6.0

Notes for Reviewers

This switches GitHub Actions workflows to the cuda13.0 branch from here: rapidsai/shared-workflows#413

A future round of PRs will revert that back to branch-25.10, once all of RAPIDS supports CUDA 13.

What about XGBoost?

This proposes the following TEMPORARILY for CUDA 13:

• not building rapids-xgboost packages
• rapids metapackage depending on CPU builds of xgboost from conda-forge

This will allow us to start publishing CUDA 13 nightlies of the rapids metapackage and other things that depend on it, like the RAPIDS container images (rapidsai/docker#782).

[jameslamb]

The test failures like this are expected:

"error": "LibMambaUnsatisfiableError: Encountered problems while solving:\n - package rapids-25.10.00a6-cuda12_py310_250828_8ca32656 requires cuda-version >=12,<13.0a0, but none of the providers can be installed\n\nCould not solve for environment specs\nThe following packages are incompatible\n\u251c\u2500 cuda-version =13.0 * is requested and can be installed;\n\u2514\u2500 rapids =25.10 * is not installable because it requires\n \u2514\u2500 cuda-version >=12,<13.0a0 *, which conflicts with any installable versions previously reported.",

(test-conda-nightly env link

Those tests only work on already-published packages, so there's always a need for an admin merge during migrations like this.

That's intentional: #770 (comment)

In other words... once we admin-merge this, that test should start passing in CI on future PRs.

[jameslamb]

Having CPU xgboost is better than no xgboost at all, I think, for anything that was relying on the rapids package to pull in xgboost.

I'm thinking this might matter for cuML notebook testing in https://github.com/rapidsai/docker, for example

[jameslamb]

/ok to test