Reproducing failing Python CI runs does not work

I have a failing Python CI run on one of my PRs: https://github.com/NVIDIA/cccl/actions/runs/16314532468/job/46077941314?pr=4847. It tells me at the end that I can reproduce the issue by running:
```
.devcontainer/launch.sh -d -c 12.9 -H gcc13 -- "./ci/test_cuda_cccl_parallel_python.sh" -py-version "3.10"
```
So I in the root directory of my CCCL clone, I run:
```
pwd
/home/bgruber/dev/cccl
gh pr checkout 4847
...
.devcontainer/launch.sh -d -c 12.9 -H gcc13 -- "./ci/test_cuda_cccl_parallel_python.sh" -py-version "3.10"
...
```
And it eventually fails with:
```
ls: cannot access '/home/coder/cccl/wheelhouse/cuda_cccl-*.whl': No such file or directory
```
Full log attached:
[log.txt](https://github.com/user-attachments/files/21255919/log.txt)