Test: cts-scheduler: add test for preventing a leftover pending monitor from causing unexpected stop of other instances

• Monitor of stateful-1 failed on node-1 so that the pending monitor didn't get overridden.
• While stateful-1 was stopping on node-1, node-2 was rejoining and pending.
• stateful-1 was stopped on node-1 and scheduler would unexpectedly decide to stop the promoted instance on node-3.