This PR, prepared as part of the work done by the AI R&D team at J‑Labs, improves dependency resolution in kubectl graph by adding several missing relationships between workloads, configuration objects, and storage resources.

The main fix is for pod configuration dependencies. kubectl graph now adds Pod -> ConfigMap relationships for:

• spec.volumes[].configMap.name
• spec.volumes[].projected.sources[].configMap.name
• envFrom[].configMapRef.name
• env[].valueFrom.configMapKeyRef.name

It also adds the analogous Pod -> Secret relationships for:

• spec.volumes[].secret.secretName
• spec.volumes[].projected.sources[].secret.name
• envFrom[].secretRef.name
• env[].valueFrom.secretKeyRef.name

In addition, pod identity and registry-related dependencies are now included:

• Pod -> ServiceAccount via spec.serviceAccountName
• Pod -> Secret with the ImagePullSecret edge label via spec.imagePullSecrets[].name

Storage relationships were extended as well:

• Pod -> PersistentVolumeClaim via spec.volumes[].persistentVolumeClaim.claimName
• PersistentVolumeClaim -> PersistentVolume via spec.volumeName
• PersistentVolume -> StorageClass via spec.storageClassName
• PersistentVolumeClaim -> StorageClass via spec.storageClassName

To make these references reliable, graph construction now pre-registers all requested resources as nodes before running resource-specific relationship analysis. This removes input-order sensitivity, so references resolve correctly even when the referenced objects appear later in the requested resource list.
When a referenced resource already exists in the graph, relationships point to the real node and preserve metadata such as labels and annotations. Otherwise, a minimal fallback node is created from kind, namespace, and name.