Add nodeSelector(nodegroup: nd) to all components, fix PriorityClass mismatch

- Volcano: set default_ns.nodegroup: nd for all components
- Kubeflow Trainer: set manager.nodeSelector.nodegroup: nd
- JobSet: set controller.nodeSelector.nodegroup: nd
- Fix PodGroup priorityClassName: normal-priority -> training-priority

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

configs/kubeflow-trainer-values.yaml

@@ -6,6 +6,9 @@
 jobset:
   install: true
   fullnameOverride: jobset
+  controller:
+    nodeSelector:
+      nodegroup: nd
 
 image:
   registry: ghcr.io
@@ -14,6 +17,8 @@ image:
 
 manager:
   replicas: 1
+  nodeSelector:
+    nodegroup: nd
   resources:
     limits:
       cpu: 200m

configs/volcano-trainjob-integration.yaml

@@ -54,7 +54,7 @@ spec:
   # MinMember ensures all pods are scheduled together (gang scheduling)
   minMember: 2
   queue: training-queue
-  priorityClassName: normal-priority
+  priorityClassName: training-priority
   minResources:
     cpu: "32"
     memory: "128Gi"

configs/volcano-values.yaml

@@ -45,6 +45,10 @@ custom:
           binpack.resources: nvidia.com/gpu
           binpack.resources.nvidia.com/gpu.weight: 5
 
+  # Node selector - schedule Volcano components on non-GPU nodes
+  default_ns:
+    nodegroup: nd
+
   # API rate limits
   scheduler_kube_api_qps: 2000
   scheduler_kube_api_burst: 2000