repo-server playbook does not distribute keys correctly when playbook is run with --limit

This patch https://github.com/openstack/openstack-ansible-repo_server/commit/9d21a73fb2244258aa4a751d1afa5b9d2faddf86 makes it possible to run the repo_server playbook with --limit without error.

There is however an unintended side effect, that keys are then only collected from hosts within the limit. This means that when redeploying a failed/reinstalled node with --limit the ssh keys from the repo master will not be installed into rebuilt repo containers. The play will succeed but subsequent repo synchronisation to the new container will fail due to missing keys.