[Feature]: Allow dcgm-exporter custom env DCGM_EXPORTER_KUBERNETES

[rtsui-harmonicinc]

Currently DCGM_EXPORTER_KUBERNETES=true is hardcoded as for the dcgm-exporter ds being generated by clusterpolicy.

This caused issue in Prometheus where timesliced pods caused metrics to be splitted into mutiple series.
NVIDIA/dcgm-exporter#340 (comment)

We wish the field can be changed as an option in clusterPolicy

[rtsui-harmonicinc]

Relevant file: https://github.com/NVIDIA/gpu-operator/blob/main/manifests/state-dcgm-exporter/0700_daemonset.yaml

[rahulait]

@rtsui-harmonicinc have you tried overriding default via clusterpolicy? So something like:

apiVersion: nvidia.com/v1
kind: ClusterPolicy
spec:
  dcgmExporter:
    env:
      - name: DCGM_EXPORTER_KUBERNETES
        value: "false"

[rtsui-harmonicinc]

thanks @rahulait , your code worked