scripts/create-kind-cluster: use nvidia-ctk stable channel & --config-source=command cli arg

[jgehrcke]

Without this patch, containerd starts up with an erroneous runc config and systemctl status containerd shows

Feb 04 13:15:38 k8s-device-plugin-cluster-worker containerd[1278]:
    time="2025-02-04T13:15:38.716063472Z" level=warning
    msg="failed to load plugin io.containerd.grpc.v1.cri"
    error="invalid plugin config: no corresponding runtime configured in `containerd.runtimes` for `containerd` `default_runtime_name = \"runc\""

We debugged this yesterday and @elezar recommended to switch to the stable release channel for the toolkit, and to use the --config-source=command arg when invoking nvidia-ctk runtime configure.

Also see NVIDIA/nvidia-container-toolkit#686.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[ArangoGutierrez]

/ok-to-test

[jgehrcke]

You live, you learn. By merging things.