Handle machines that stay in allocating + pending for too long.

[freyes]

There are scenarios where the undercloud infrastructure could have issues with the metadata service, this results in a machine that's running, but since it couldn't be configured by the scripts juju passes in cloud-init, the juju status will show it as "pending" for the machine status and "allocating" for the agent status, zaza should monitor this condition and short circuit the testing if this state stays for too long.

this functionality should include a configuration like TEST_MACHINE_ALLOCATION_TIMEOUT.