wheels: build CUDA 13 wheels with latest CTK (13.3.0)

[jameslamb]

Contributes to rapidsai/build-planning#268

Part of a RAPIDS-wide initiative to build wheels using v13.3.0 of the CUDA Toolkit, and to generally get RAPIDS back to the pattern of always building against its latest supported CTK.

As of rapidsai/shared-workflows#590, RAPIDS builds CUDA 13 wheels with v13.3 of the CTK. This updates the project's nvidia-nvjitlink floor to match that.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[jameslamb]

/ok to test

[jameslamb]

2 wheel test jobs timed out after 6 hours, though one was a CUDA 12.9 job so I suspect it's not related to this PR:

https://github.com/NVIDIA/cuvs/actions/runs/28916932499/job/85792220262?pr=2305

[jameslamb]

If we get to the point where only conda-cpp-tests (which can take 2+ hours) is still running and everything else has passed, I'll admin-merge this.

All CI will run here because I touched dependencies.yaml, but these changes have 0 effect on anything outside of wheels.

And I want this merged + new packages built on main before merging rapidsai/cugraph#5575, to hopefully get out new cuGraph builds built with new-enough RAFT and cuVS ... in cugraph-gnn, we are still dealing with the fallout from the RAFT ABI break a few days ago, I think because a new-enough build of that entire combination hasn't yet been published: rapidsai/cugraph-gnn#487

[jameslamb]

Only conda-based CI is left here, and this change wouldn't affect it at all. Cancelling CI to save resources, admin-merging to keep the migration moving.