Instance-terminate workflow reliable activation by using taskflow

After moving the core logic of the terminate_instance 'workflow' to the conductor (and leaving a lightweight task-receiver in the compute manager that still runs driver like logic) we need to be able to reliably ensure that that workflow is executed in a manner that can survive process failure by resuming (or can be reverted in this case); this will be useful for upgrades, crashes, and other general cases where reliability and consistency of your cloud is important (when is it not). Taskflow provides this type of core functionality so it seems like an ideal fit for this scenario.