This test was designed with out the notion that it could be running in a cluster with TNA and workers, as such it can fail sometimes when the pod is created on one of the workers. We need to update the test to signal success when pod runs on a worker.

Failed RUN: https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/openshift_release/71398/rehearse-71398-periodic-ci-openshift-release-master-nightly-4.21-upgrade-from-stable-4.20-e2e-metal-ovn-two-node-arbiter-upgrade-workers/1998490809527177216

Test assumes no workers, thus breaking here: https://github.com/openshift/origin/blob/737f7384774e6274ba48cd74392c9f4564889488/test/extended/two_node/arbiter_topology.go#L185-L186

Acceptance Criteria:

Invert the logic so we check that we pull the arbiter node name, and check that the pod is not on the arbiter, master/workers are assumed

https://github.com/openshift/origin/blob/737f7384774e6274ba48cd74392c9f4564889488/test/extended/two_node/arbiter_topology.go#L151-L153