[JENKINS-47821] Prevent run-once slave from accepting more jobs.

[pjdarton]

vSphereCloudSlaveTemplate

• now provides each cloud slave it creates with its own RetentionStrategy instance, being a copy of the template's rather than being the same instance. This allows us to simplify RunOnceCloudRetentionStrategy's implementation as it's no longer shared.

RunOnceCloudRetentionStrategy

• can no longer be shared between slaves. It now only holds state information about one slave. See also: vSphereCloudSlaveTemplate changes.
• now overrides isAcceptingTasks to prevent further activity instead of simply setting c.setAcceptingTasks(false).
• now triggers node termination in-line instead of spawning a new thread for it. We can do this without slowing Jenkins down because our nodes ask vSphere to delete their VMs in a separate thread. Hopefully
  this should fix the re-use bug.
• offline reasons now given as a VSphereOfflineCause that can be localized. See also: vSphereCloudLauncher changes.

vSphereCloudSlave

• flyweight-task refusal message can now be localized.
• offline reasons now given as a VSphereOfflineCause that can be localized. See also: vSphereCloudLauncher changes.

vSphereCloudLauncher

• no longer uses the special text "vSphere Plugin" in the offline reason to recognise our own offline reasons. Our code uses now VSphereOfflineCause and so we can now use instanceof to detect these
  instead of String.contains.

[pjdarton]

I've been running a plugin with these changes in for a week now and it all seems fine to me.
At the very least, these changes are harmless.
Hopefully they've actually fixed something.