Fix Pod label, Nodegroup and nvidia chart name issue

[Micky-Yang]

In my actual reference documents, during the practise, I found the following problems:

1. Helm chart repository name conflict related to Nvidia

helm repo add nvidia https://nvidia.github.io/k8s-device-plugin
helm repo add nvidia https://helm.ngc.nvidia.com/nvidia
helm repo add nvidia https://nvidia.github.io/gpu-operator

All helm repos are named nvidia, which causes repo name conflicts. Different repo names have been used to distinguish them.

2. Automatic installation instructions for the Nvidia device plugin service

3. pending issues of coredns and metrics-server services
   The Pod cannot be scheduled because the default gpu-dra-nodes adds taints:

kube-system   coredns-7bf648ff5d-4bs45               0/1     Pending   0          12m
kube-system   coredns-7bf648ff5d-n4bwq               0/1     Pending   0          12m
kube-system   metrics-server-7fb96f5556-4cpdl        0/1     Pending   0          12m
kube-system   metrics-server-7fb96f5556-6mbvh        0/1     Pending   0          12m

So a base-nodes NodeGroup was added to fix this problem.

4. Pod Pending issue caused by mismatch between Pod nodeSelector label and NodeGroup label key

The Pod nodeSelector label is NodeGroupType: gpu-dra, but NodeGroup label is node-type: "gpu-dra", so change the Pod nodeSelector label to node-type: "gpu-dra".