Persistent volumes are created dynamically, w/ default storageClass

[DougReeder]

hostPath volumes are labeled a "dangerous escape hatch" in the documentation, and will be discarded if there is ever a second node. There are several mechanisms where this may happen, even for a "single-node" cluster, such as when the version of Kubernetes is upgraded by spinning up a new node, then draining and spinning down the old.

This PR creates volumes

1. off the node, so they outlast nodes
2. using the cluster's default storage class. This is either the provider's default (which is some standard/general purpose flavor of storage for every provider I checked), or was specifically set by the cluster's administrator.

Cloud Provider	Default StorageClass Name	Default Provisioner
Amazon Web Services	gp2	aws-ebs
Microsoft Azure	standard	azure-disk
Google Cloud Platform	standard	gce-pd
OpenStack	standard	cinder
VMware vSphere	thin	vsphere-volume
DigitalOcean	do-block-storage	dobs.csi.digitalocean.com
Scaleway	sbs-default	csi.scaleway.com

3. that can be backed-up and restored on a different provider
4. can be expanded (usually)

Currently, DigitalOcean charges $1/mo for a 10GB volume, so this PR adds $2/mo to a default installation.

See my blob post for background.

It also creates the pgsql volume with accessMode ReadWriteOncePod. If someone ever tries to scale up the pgsql deployment, this will ensure only one PostgresQL engine can access the files, so they aren't corrupted. (ReadWriteOnce allows all pods on the node to access the volume.)

It also tweaks the mountPath for PostgresQL so it's not creating files and directories at the root of the volume, so PostresQL doesn't balk at the presence of a lost+found directory.

It also removes the duplicate Secret.

This PR has been tested on Digital Ocean; it would be desirable to validate it on another provider.

[DougReeder]

@Exairnous has pointed out this is a breaking change. I think the best way to deal with this is a flag in input-values.yaml that defaults to hostPath volumes. Then admins can migrate to external storage independently of update to the latest Hubs implementation. That also provides a "budget" option, which is also useful for test clusters.