fix(ci): use nvidia RuntimeClass instead of pod_spec for GPU access

Replace pod_spec strategic merge patch (silently not applied) with
runtime_class_name="nvidia". The nvidia RuntimeClass uses the
nvidia-container-runtime which injects GPU drivers, nvidia-smi, and
/dev/nvidia* devices into all containers automatically.

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

infra/k8s/gitlab/runner-values.yaml

@@ -37,6 +37,8 @@ runners:
         helper_memory_request = "128Mi"
         helper_memory_limit = "512Mi"
         image_pull_secrets = ["scw-registry", "gitlab-registry"]
+        # Use nvidia RuntimeClass — injects GPU drivers + /dev/nvidia* via nvidia-container-runtime
+        runtime_class_name = "nvidia"
       # Sub-tables must come AFTER all scalar values (TOML rule)
       [runners.kubernetes.node_selector]
         "k8s.scaleway.com/pool-name" = "gpu-training"
@@ -44,11 +46,6 @@ runners:
         "nvidia.com/gpu" = "NoSchedule"
       [runners.kubernetes.pod_labels]
         "app.kubernetes.io/part-of" = "foxhunt-ci"
-      # Request GPU so NVIDIA device plugin injects /dev/nvidia* into build pods
-      [[runners.kubernetes.pod_spec]]
-        name = "build"
-        patch_type = "strategic"
-        patch = '{"spec":{"containers":[{"name":"build","resources":{"limits":{"nvidia.com/gpu":"1"},"requests":{"nvidia.com/gpu":"1"}}}]}}'
 
   # Runner tags for job matching
   tags: "kapsule,rust,docker,gpu"