This commit introduces changes to try harder to prevent the following error from appearing:
```
2020/11/25 15:53:58 [ERROR] API error response 500 for POST /v3/clusters/c-l4bdt?action=enableMonitoring. Cause: Operation cannot be fulfilled on clusters.management.cattle.io "c-l4bdt": the object has been modified; please apply your changes to the latest version and try again
```

The root cause of this error is that we do not normally retry on a conflict, so before it was appearing even after spinning up just 4 clusters in succession and trying to immediately enable monitoring.

Related Issue: https://github.com/rancher/rancher/issues/30188

Signed-off-by: Arvind Iyengar <arvind.iyengar@rancher.com>
(cherry picked from commit fa66cced)

[release/v2.5] Aci cni backport

Add support for startTLS

Signed-off-by: Kiran Shastri <shastrinator@gmail.com>
(cherry picked from commit 0af5ba82)

Signed-off-by: Kiran Shastri <shastrinator@gmail.com>

Signed-off-by: Kiran Shastri <shastrinator@gmail.com>

[2.5] Updated chart and tests to allow specify ember and vue Rancher custom logos persistence

[v2.5] add mirantis check + typos/grammar

[2.5]Backport-Add ability to set password for rancherd

[2.5] Update provider detector to include rke.windows

Adds options to global DNS [2.5]

[2.5] Backport: Update rancherd to point to rke2 v1.18.12+rke2r1

[forwardport] Move creation of accesscontrol

Problem:
When a cluster agent is diconnected rancher tries to start the
controllers in a slow loop, but this causes a memory leak. The leak
stems from adding a handler in client-go which has no method to
deregister the handler so a fixed size array leaks every time.

Solution:
Don't create access control until Rancher has verified a connection to
the cluster which is a good indication controllers will start. There is
technically a small leak here every time we start these controllers.

[2.5] Backport http proxy fleet

Co-authored-by: Dan Ramich <dramich@users.noreply.github.com>

[Backport] 2.5: Cluster name option v2.5

[Backport] v2.5: Fix version panic v2.5

Added nil check for version on status. Prior, version could be
accessed before it was set. This would occur on clusters that were
newly provisioned or in the process of being provisioned and did
not have version set yet. Now, if the version is not set an error
will be returned so that controllers can retry until version is
available.

[2.5] Add rollback actions test note for 2.4 and below

[2.5] Handle rollback actions

Co-authored-by: Thijs van Son <thijs.van.son@kvk.nl>
Co-authored-by: stayfrostnl <thijsvanson@live.nl>
Co-authored-by: Thijs van Son <thijs.van.son@kvk.nl>

Add apiVersion changes back in

[Backport]2.5: Fix return err when secret webhook-receiver is not found

[2.5 Backport] Update steve vendor to include exec fix

* Prevent getting cluster.name when clusterClient returns an error

* wrap rancherd reset-admin errors

[2.5] Make security groups optional when VPC is provided

[2.5] Fix test_writing_config_to_disk timeout issues