New cluster may put Waypoint server on nginx node

[ragesoss]

The node sizes are designed so that nginx is the only job running on its node, but I just spun up a fresh server and the waypoint server ended up being allocated to the nginx node first, so there wasn't enough memory left to put the nginx node on it. I guess we need some constraints to make sure waypoint server and runner end up on a generic node instead of the nginx node.

[ragesoss]

Related to this, the latest version of waypoint (0.6.1) looks like it requires a nomad-host-volume to be specified. (I guess this provides persistence across server installs that was causing problems before.) I think we should adjust the configuration to include a persistence volume on every node (including the novol nodes, node-0 and node-1, that currently lack them, and then use the node-0 volume for the waypoint install. That will hopefully also ensure that the waypoint-server and waypoint-runner jobs end up running on node-0 and not the nginx node.