Test the cluster upgrade support of the GPU Operator about ci-artifacts

[kpouget]

I have started testing how OpenShift upgrade behaves with this PR on the CI repository: openshift/release#17498 and using the release-next branch to control the code running in the CI.

My idea is to first test:

1. Install and test the GPU Operator
2. Uninstall the GPU Operator
3. Upgrade
4. Reinstall and test

then we'll skip step 2 when the seamless upgrade is merged&published.

[kpouget]

The cluster upgrade support has been merged in the GPU Operator master branch (will be part of release 1.7.0 due within next week), see PR 211 and PR 214.

In this repo, I added a role to trigger the cluster upgrade #107 and added the ability to test the GPU Operator deployment without deploying it: #125

And this morning, we got the first proper upgrade test to pass!

Last steps of this work are:

• properly integrate the test in Prow configuration,
• decide a testing frequency and
• add an entry in the GPU Operator CI Dashboard

[kpouget]

PR openshift/release#18146 finished and passing, only waiting for \lgtm for merging