skip CUDA 11.4 conda-python-tests jobs (#91)

CUDA 11.4 is only tested in nightly CI (not on PRs) for this project.

This project doesn't support CUDA 11.4 (because the versions of PyTorch
it is compatible with don't), so #89 turned off most of those CUDA 11.4
test jobs. That worked for wheels, but conda tests need to be turned off
as well.

```text
tests/data/test_dask_graph_store.py::test_get_edge_index[multi_edge_multi_vertex_graph_1-cudf] - AssertionError: Torch not compiled with CUDA enabled
```

([build
link](https://github.com/rapidsai/cugraph-gnn/actions/runs/12198233986/job/34029537082#step:8:141493))

This skips nightly tests against CUDA 11.4.

## Notes for Reviewers

### How I tested this

Triggered a run of the nightly test job from this branch ([by clicking
"Run workflow"
here](https://github.com/rapidsai/cugraph-gnn/actions/workflows/test.yaml)),
using the latest nightly packages:

* **use workflow from:** `pull-request/91`
* **branch:** `branch-24.12`
* **date:** `2024-12-06`
* **sha:** `7ec8ace3cf7007dcca796e484f587af77d328e8f` (from
https://github.com/rapidsai/cugraph-gnn/commits/branch-24.12/)

Build link:
https://github.com/rapidsai/cugraph-gnn/actions/runs/12203424243/job/34046336319

Author: jameslamb

.github/workflows/test.yaml

@@ -40,6 +40,7 @@ jobs:
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
+      matrix_filter: map(select((.ARCH == "amd64") and (.CUDA_VER | startswith("11.4") | not)))
   wheel-tests-pylibwholegraph:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]