[bug] gk-deploy script with our pinned version fails on K8s 1.10

[leifmadsen]

With the version of Heketi we have pinned, deployment to master will fail unless you revert your K8s version back to 1.9.6 or earlier since the console output seems to have changed just enough on 1.10 that the check_pods() function is unable to determine that the glusterfs pods have spun up.

Current work around is to pass -e "kube_version=1.9.6-0" to your ansible-playbook run so that K8s 1.10 isn't installed.

Fix is probably to bump our Heketi deployment pin, which we've been meaning to do for a while per #165