Test: scheduler: Add tests for failed probe operations.

This adds identical sets of tests for primitive resources and cloned
resources. For the moment, the output reflects the current state of the
code. No changes have been made to properly handle these operations
yet.

Each set has three resources, and each is set up with a slightly
different configuration of probe failures:

(1) - Maskable probe failure on each node.
(2) - Maskable probe failure on one node, successful "not running" probe

on the other node.  The resource should be started on the node
where "not running" was returned.

(3) - Maskable probe failure on one node, non-maskable probe failure on

the other node.  The resource should not be running anywhere, and
should be stopped on the node with the non-maskable failure.