Large merge-back of soak-cluster-deployers

[MrMarvin]

Goal of this PR

Providing an idempotent, up-to-date and reliable set of well documented Ansible roles to automate the full life-cycle of Mesosphere DC/OS.

Things to consider

• Dropping everything but the Ansible stack in this repository. Infrastructure definitions should be taken from the official Mesosphere DC/OS terraform stack.
• Only well tested for DC/OS 1.12
• Only supports CentOS 7
• This PR is huge. I'd recommend to not go through each file, but rather inspect the state as-is in the source branch and decide from there if the architecture makes sense to be used for dcos-ansible.

Features

• Fully automated DC/OS installation
• Fully (or semi-, if desired) automated DC/OS upgrades
• Automatic 'same version config updates' whenever the generated DC/OS config changes on bootstrap
• Upgrade on same version, using the image_commit over version. Required for upgrading between non released versions.
• Multiple roles, sharing code for DC/OS prerequisites in an additional role
• Fully idempotent, single playbook
• All variables securely name spaced (e.g. dcos['...'])
• Not relying on inventory groups, using group_vars instead for usage in existing projects
• Supports all current and future DC/OS config parameters via config.yaml pass-through.
• Documentation and examples

[MrMarvin]

As a preview on how this would end up in the Ansible Galaxy, I tweaked my personal Github fork to point to the feature branch per default (instead of master), added some Galaxy meta data files to the individual roles and created a new v0.42.2 git tag. The result is here: https://galaxy.ansible.com/mrmarvin/ansible_dcos
(Hopefully the transition pulling this to another Galaxy name space once this is merged is as easy...)

[jrx]

Working on testing/integrating this PR in:
https://github.com/dcos-labs/ansible-dcos/tree/MrMarvin-soak-cluster-deployers-merge-back