[CP 1295] feat(tests/k8s-e2e): add GPU Operator e2e test suite with DME verific…

[ci-penbot-01]

cp of pensando/gpu-operator#1295

---

Source PR Description (pensando/gpu-operator#1295):

…ation

Adds a containerized GPU Operator e2e test suite (Op000-Op900) that covers the full operator lifecycle with DME integration:

• Op000: pre-flight checks
• Op010-Op020: operator install and readiness
• Op030: KMM controller verification
• Op040: DME DaemonSet verification
• Op050: GPU health check
• Op060: metrics endpoint validation
• Op065: partition configuration validation
• Op070: GPU workload (PyTorch matmul)
• Op900: operator teardown

Includes Dockerfile for containerized test runner.

2026/04/06 01:07:03 Op000: adding cert-manager helm repository
  2026/04/06 01:07:03 Op000: installing cert-manager
  2026/04/06 01:08:46 Op000: cert-manager installed
  2026/04/06 01:08:46 Op001: installing GPU Operator main-78a1f8d9 (DME: 1.5.0-rocm7.12-caa3d592)
  2026/04/06 01:08:52 Op001: GPU Operator installed
  2026/04/06 01:08:52 Op010: waiting for NFD to label node feature.node.kubernetes.io/amd-gpu=true
  2026/04/06 01:09:03 Op010: 1 node(s) have amd-gpu label
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.device-id=75a8
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.product-name=AMD_Radeon_Graphics
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.family=AI
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.driver-version=6.18.8
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.vram=144G
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.simd-count=512
  2026/04/06 01:09:03 Op010: node galena3836 label amd.com/gpu.cu-count=128
  2026/04/06 01:09:03 Op020: node galena3836 allocatable amd.com/gpu = 1
  2026/04/06 01:09:03 Op030: found 1 KMM controller pod(s); phase=Running
  2026/04/06 01:09:03 Op040: using DME pod default-metrics-exporter-w52mr (phase=Running)
  2026/04/06 01:09:08 Op040: DME returned 98 metric families; all required metrics present
  2026/04/06 01:09:13 Op050: gpu_health — healthy=1 unhealthy=0
  2026/04/06 01:09:19 Op060: category "ecc" OK
  2026/04/06 01:09:19 Op060: category "clock" OK
  2026/04/06 01:09:19 Op060: category "vram" OK
  2026/04/06 01:09:19 Op060: category "pcie" OK
  2026/04/06 01:09:19 Op060: category "power" OK
  2026/04/06 01:09:19 Op060: category "temperature" OK
  2026/04/06 01:09:19 Op060: 98 total metric families verified
  2026/04/06 01:09:24 Op065: detected 1 partitioned GPU instance(s)
  2026/04/06 01:09:24 Op065: GPU 0 partition_id=0 compute_type=SPX memory_type=NPS1
  2026/04/06 01:09:24 Op065: per-partition metric "gpu_gfx_busy_instantaneous" present for all 1 partition(s)
  2026/04/06 01:09:24 Op065: per-partition metric "gpu_total_vram" present for all 1 partition(s)
  2026/04/06 01:09:24 Op065: per-partition metric "gpu_used_vram" present for all 1 partition(s)
  2026/04/06 01:09:24 Op065: partition validation passed for 1 GPU partition(s)
  2026/04/06 01:09:25 Op070: submitting GPU workload pod op-e2e-gpu-workload
  2026/04/06 01:09:35 Op070: workload logs:
                     ROCm available: True
                     DONE
  2026/04/06 01:10:14 Op900: teardown complete
  OK: 11 passed
  --- PASS: Test (211.66s)

Cherrypick triggered by: ACP-Automation

[spraveenio]

Lgtm

[yansun1996]

LGTM