duplicate devicePlugin config in gpu-operator

devicePlugin is defined twice (at the same level) but with different configurations for the gpu-operator in cern-magnum:

cat values.yaml | yq .gpu-operator | grep devicePlugin -A 5
devicePlugin:
  repository: registry.cern.ch/kubernetes
  version: v0.13.0
--
devicePlugin:
  repository: registry.cern.ch/kubernetes
  version: v0.12.2-ubi8
  config:
    name: nvidia-device-plugin-config
    default: "default"

helm seems to take the latest but we should cleanup for avoidance of confusion.