currently if the ACTIVE netboot-ed node with driver based on AgentDeploy
is being taken over by another conductor due to the whole host with
previous conductor died/went offline, the node will not be fully
operational as it cannot be rebooted from inside:
- its DHCP options are still pointing to the TFTP/HTTP servers of old
(now offline) conductor host
- new conductor host does not have the user image kernel and ramdisk
cached
- new conductor host lacks PXE env prepared for such node.
Fix this by actually calling out to `driver.boot.prepare_instance`
in AgentDeploy.prepare for ACTIVE nodes.
Change-Id: Ie21e285e22c6a8b168f8c9776a125f3bb71fead7
Closes-Bug: #1646038