Update nvidia docker image. This was missed last time.

[ScottTodd]

The postsubmit test_nvidia_a100 job started failing after this was merged: https://github.com/iree-org/iree/actions/runs/9493590493/job/26189543324#step:8:60

The persistent runner is out of disk space: docker: failed to register layer: write /var/cuda-repo-ubuntu2004-12-2-local/nsight-systems-2023.2.3_2023.2.3.1001-1_amd64.deb: no space left on device.

Ideas:

• Check the docker image size before and after this change
• Check the docker image cache on the runner
• Clear the docker image cache on the runner and hope multiple jobs don't redownload different images
• Increase the disk size
• See what else is taking up space on the runner (old run artifacts? install directory? benchmark model weights etc.?)

[ScottTodd]

• Check the docker image size before and after this change

12.3 GB both before and after