Make clustering scenarios cleanup more thorough

CONTROLLER-1680 and CONTROLLER-1681 already describe situations where a failure from the previous scenario affects the next scenario. The easiest and most reliable cleanup is to use the reboot suite, maybe as a test case in a longer suite.

The recent severe example is this [0] failure, when a leader movement from the previous suite is seen as Candidate state, causing the suite to skip all the relevant checks.

[0] https://logs.opendaylight.org/releng/jenkins092/controller-csit-3node-clustering-only-carbon/686/archives/log.html.gz#s1-s41-t2-k2-k3-k1-k6-k2-k2-k1