[Master]Forward Port https://github.com/rancher/rancher/issues/22790

Handle rollback actions

Add check for taints to determine if pod can run

Create reference from workload to secret

Remove basic auth for GKE

Fix helm repo constant refreshing

Match unreachable taint key without comparing effect

Problem: Till k8s 1.18.10, the k8s node lifecycle controller sets these two taints on an unreachable node
(or a node with Ready=Unknown): `node.kubernetes.io/unreachable:NoSchedule` and `node.kubernetes.io/unreachable:NoExecute`.
In k8s 1.19.3, the node controller is only setting the taint with NoSchedule effect for such a node.
The nodepool controller checks for the presence of a taint with key `node.kubernetes.io/unreachable` and effect `NoExecute`
when determining if a node is unreachable and needs to be replaced. Since the node in 1.19 does not get this taint, nodepool
controller does not consider it as unreachable.

Solution: Modify the check in nodepool controller to consider a node unreachable even if only the taint key matches
 `node.kubernetes.io/unreachable`. The taint effect should not impact how nodepool processes this node.

The helm repo was being refreshed because `repoSpec.URL` and
`status.URL` were different. They were different because we were
ensuring that `status.URL` ended with a '/'. Now we only add the slash
in the repo URL if it is needed to download a chart.

The basic was used to generate an access token. However, the access
token was also being generated when getting a service client. Instead of
using the basic auth, we now use the access token.

Adding K8s version for import clusters

vendor update rke

Show states on relationships

Disable all endpoints for local cluster if rancher agent is present.

Fix nil pointer dereference in EKS controller

The  catalog manager was not initialized with the eksOperatorController
and this caused a nil pointer dereference when it was used.

This reverts commit c5ec480f.

Ensure that non-privileged users can still get to the shell

Add kubeVersion checks for catalogv1

enqueue node if node plan isn't initialized after rke up

[master]Update provider detector to include rke.windows

Increase default timeout to avoid test failures

Change ingress spec when using newest apiVersion

Fix service accounts for clusterrepos