Refactor playbooks for discrete initialization, package update, and normal usage

• Separate playbooks for initializing everything. What are now named *_all.yml

• Separate playbook for upgrading packages. I have this already for the Rackspace servers.

It's the package upgrading that's the main concern, because you don't want that happening accidentally. The upgrade playbook needs to have a reboot option, and you wouldn't want servers being rebooted because of accidentally running the "all" playbook with that option enabled.

Package updates are too time-consuming on the old Rackspace systems, so I'm looking to minimize the effort required right now.