Introduces ExistingClusterKubeconfig option to unmanaged-cluster

[jpmcb]

What this PR does / why we need it

This PR introduces the ExistingClusterKubeconfig config option which is supported by the --existing-cluster-kubeconfig flag during create.

This enables users to bring their own clusters that already exist and already have a kubeconfig on the filesystem.

Why not just use KubeconfigPath?

This would have required a greater refactor since the config.KubeconfigPath is used to create and set the context of new created kind clusters:

clusterConfig := kindcluster.CreateWithkubeconfigPath(c.KubeconfigPath)

...

err = kindProvider.Create(c.ClusterName, clusterConfig, kindConfig)

This means that we would have been overloading the use of KubeconfigPath: it would both be the path to the kubeconfig for some arbitrary created cluster or used in logic to determine if a user is attempting to provide an existing cluster.

Instead, introducing a new option allows us to keep Kubeconfig just as it is: the file path to a kubeconfig.

Deeper in the implementation, we actually take advantage of this by setting the Kubeconfig to the what is provided by a user when using ExistingClusterKubeconfig. This enables existing program flow to stay in place without creating extra logic around a separate kubeconfig.

Details for the Release Notes (PLEASE PROVIDE)

introduce ExistingclusterKubeconfig option for unmanaged-cluster

Which issue(s) this PR fixes

Fixes: #2890

Describe testing done for PR

Existing functionality intact, still able to create cluster and target it:

• create some arbitrary cluster. Used kind for this example:

$ kind create cluster --name already-exists
...

• Get the cluster's kubeconfig:

$ kind get kubeconfig --name already-exists > /tmp/kubeconfig.yaml

• Give unmanaged-cluster the kubeconfig as an option:

Check the stats of the cluster:

❯ k get pods -A --kubeconfig kubeconfig.yaml
NAMESPACE            NAME                                                   READY   STATUS    RESTARTS   AGE
kube-system          coredns-558bd4d5db-j72jw                               1/1     Running   0          2m52s
kube-system          coredns-558bd4d5db-psrm9                               1/1     Running   0          2m52s
kube-system          etcd-already-exists-control-plane                      1/1     Running   0          2m55s
kube-system          kindnet-gx8v4                                          1/1     Running   0          2m52s
kube-system          kube-apiserver-already-exists-control-plane            1/1     Running   0          2m55s
kube-system          kube-controller-manager-already-exists-control-plane   1/1     Running   0          2m54s
kube-system          kube-proxy-8lgm2                                       1/1     Running   0          2m52s
kube-system          kube-scheduler-already-exists-control-plane            1/1     Running   0          2m55s
local-path-storage   local-path-provisioner-547f784dff-rfhv8                1/1     Running   0          2m52s
tkg-system           kapp-controller-f7dbbcb4f-nrpbz                        1/1     Running   0          89s

Special notes for your reviewer

N/a

[joshrosso]

exciting! putting this blocker in so i remember to review/validate today.

[joshrosso]

Validated this works great against Rancher Desktop.

I also added an update to support --cni=none, which I forsee being common for some of these existing cluster deployments.

Ready to approve once we remove the old KubeconfigPath field.