docs: Describe device list settings when CDI is enabled

[mikemckiernan]

Review HTML: https://nvidia.github.io/cloud-native-docs/review/pr-525/gpu-operator/latest/cdi.html

Summary

Adds a "CDI and Device List Settings" subsection to gpu-operator/cdi.rst, and a cross-reference to it from "Enabling CDI". The subsection tells administrators that, when CDI is enabled:

• They don't need to set DEVICE_LIST_STRATEGY. The Operator sets cdi-cri,cdi-annotations on the device plugin, and cdi-annotations is the fallback when CRI CDI support is missing.
• If they set devicePlugin.env DEVICE_LIST_STRATEGY themselves, that value replaces the Operator's value. The value must contain only cdi-cri,cdi-annotations, so volume-mounts and envvar must be removed.
• They must remove toolkit.env ACCEPT_NVIDIA_VISIBLE_DEVICES_AS_VOLUME_MOUNTS.

The section names the symptom (unresolvable CDI devices management.nvidia.com/gpu=GPU-<uuid>) and includes a valid YAML example.

Basis: transformDevicePluginCtrForCDI in gpu-operator controllers/object_controls.go, and the repro in NVIDIA/gpu-operator#3077. Related: NVIDIA/gpu-operator#3081 (guardrail, open).

The local gpu-operator build produced no new warnings; the existing tab-set errors in dra-intro-install.rst and gpu-driver-upgrades.rst are unrelated. The pre-commit hooks could not be fetched in the authoring environment; I checked trailing whitespace by hand.

Questions for reviewers

1. Does volume-mounts fail even when ACCEPT_NVIDIA_VISIBLE_DEVICES_AS_VOLUME_MOUNTS is unset?
2. Do pods that don't set runtimeClassName: nvidia fail the same way?
3. If the default management.nvidia.com/gpu kind is fixed for workload pods (gpu-operator#3077), does this guidance change?
4. If gpu-operator#3081 merges, should the section name the ClusterPolicy error?
5. Does the same guidance apply to GPUCluster?
6. Does the device plugin choose between cdi-cri and cdi-annotations at run time? The fallback sentence depends on the answer.

Link to Devin session: https://nvidia-cloud.devinenterprise.com/sessions/9c26f672c2354d3e88fe31b5eb47c065
Open in Devin Desktop: https://nvidia-cloud.devinenterprise.com/desktop/session/9c26f672c2354d3e88fe31b5eb47c065?variant=devin
Requested by: @mikemckiernan

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[github-actions]

Documentation preview

https://nvidia.github.io/cloud-native-docs/review/pr-525