In openshift_repos and everywhere, ensure deployment_type and
openshift_deployment_type are defined and the same.
We really want to set openshift_deployment_type, but users will likely
still have just deployment_type, so accept both. And don't make every
playbook default openshift_deployment_type to deployment_type.

This introduces the openshift_sanitize_inventory role to run before anything else.

- Add openshift_facts role and module
  - Created new role openshift_facts that contains an openshift_facts module
  - Refactor openshift_* roles to use openshift_facts instead of relying on
    defaults
  - Refactor playbooks to use openshift_facts
  - Cleanup inventory group_vars

- Update defaults
  - update openshift_master role firewall defaults
    - remove etcd peer port, since we will not be supporting clustered embedded
      etcd
    - remove 8444 since console now runs on the api port by default
    - add 8444 and 7001 to disabled services to ensure removal if updating

- Add new role os_env_extras_node that is a subset of the docker role
  - previously, we were starting/enabling docker which was causing issues with some
  installations
  - Does not install or start docker, since the openshift-node role will
    handle that for us
  - Only adds root to the dockerroot group
  - Update playbooks to use ops_env_extras_node role instead of docker role

- os_firewall bug fixes
  - ignore ip6tables for now, since we are not configuring any ipv6 rules
  - if installing package do a daemon-reload before starting/enabling service

- Add aws support to bin/cluster

- Add list action to bin/cluster

- Add update action to bin/cluster

- cleanup some stray debug statements

- some variable renaming for clarity

- Rename repos role to openshift_repos
  - Make openshift_repos a dependency of openshift_common
  - Add README and metadata for openshift_repos
  - Playbook updates for role rename
- Verify libselinux-python is installed, otherwise some of the bulit-in
  modules we use fail