Add GPUClusterConfig CRD and controller for DRA-based stack

[karthikvetrivel]

1. Overview

We introduce a new CRD named GPUClusterConfig and a new controller for reconciling it. Like ClusterPolicy today, it is a singleton, cluster-scoped CRD that configures the operands needed to enable GPUs in Kubernetes. GPUClusterConfig represents the new DRA-based software-enablement stack; it is an evolution of ClusterPolicy.

Change Log

9f08bec:

• Defined GPUClusterConfig Go types in api/nvidia/v1alpha1, cluster-scoped + singleton, with kubebuilder validation/default markers for every operand block. Wire AddToScheme. Generated the CRD manifest + deepcopy.
• Tested: make manifests generate produces the CRD yaml and deepcopy. kubectl apply the CRD succeeds.

73c9d30:

• Introduced a new controller built on the existing state.Manager / SyncState() engine (the same
  pattern NVIDIADriver uses), registered in cmd/gpu-operator/main.go.
• Singleton enforcement (first-wins): a single instance owns reconciliation; any additional
  instance is marked Ignored and skipped. Mirrors how ClusterPolicy handles duplicates.

ccc0f7a:

• Added cmd/dra-driver-validator, the init container binary for the DRA kubelet-plugin DaemonSet. It runs before the gpus and computeDomains containers start, validates that the NVIDIA driver is installed, and writes /run/nvidia/validations/driver-ready with the two env vars the kubelet-plugin containers source on startup (NVIDIA_DRIVER_ROOT, DRIVER_ROOT_CTR_PATH).
• Tested: unit tests end-to-end with fake driver, and against a real NVIDIA 595.58.03 driver on an A100 node.

87fa6c0:

• Adds the DRA driver operand to the GPUClusterConfig controller (a new state-dra-driver state with its manifests wired into the state manager)