Ensure that requested instances are actually created

The current implementation of nova handles instance creation in an asynchronous manner in order to prevent the sort of blocking that would limit the scalability of the system. This is achieved by casting VM creation requests onto the queue, and assuming that they will eventually be completed successfully. As this is not an accurate assumption, we need to create a means to follow up on these requests, and in the event that one or more requested instances failed to be successfully created, have the option to re-issue the create request for those failed instances.