Draft: Repair cluster-build pipeline

The introduction of building RHEL8 images on the cluster in MR !1681 (merged) had a flaw: it would recreate image streams on every CI run, thereby losing the association to images in the image registry. In the CI, we change from time to time the ran-base image between tags develop and ci-temp. By recreating the image streams, we quickly lost the association of ran-base:develop, which the CI assumed to "just be there". In this case, most builds would fail, basically bringing the CI to a halt. The actually important change here is in 83736340

Marking as BUILD-ONLY as we cannot test this in the CI: the CI will trigger the normal RHEL8 build on asterix. I will reproduce the scenario we had previously, i.e., trigger this pipeline multiple times in a way that builds will change the ran-base image between ran-base:develop and ran-base:ci-temp. The expected outcome is to always have a successful build, of course.

@lthomas I have additional changes to not use the SSHConnection class, but this will come later. The same with the cleanup of the physims.

ci-scripts/cls_cluster.py

@@ -255,8 +255,8 @@ class Cluster:
 
 		status = True # flag to abandon compiling if any image fails
 		attemptedImages = []
-		self._recreate_is_tag(mySSH, 'ran-base', baseTag, 'openshift/ran-base-is.yaml')
 		if forceBaseImageBuild:
+			self._recreate_is_tag(mySSH, 'ran-base', baseTag, 'openshift/ran-base-is.yaml')
 			self._recreate_bc(mySSH, 'ran-base', baseTag, 'openshift/ran-base-bc.yaml')
 			ranbase_job = self._start_build(mySSH, 'ran-base')
 			attemptedImages += ['ran-base']