Per-node disk paths not intuitive to users

As per some training feedback, it is not immediately intuitive to users that their non-default disk paths for indexes and data are not propagated to nodes as they are added to the cluster. This is somewhat by design so that the paths can be set per-node, but users are not familiar with this.

In order of ease, I would suggest a few possible improvements:
1 - Update the documentation simply to state the best practices here that you need to set the disk paths for a node before adding it to the cluster. This is minimal effort, but not necessarily solving the problem if users don't read every line of the docs.
2 - Add a message on the "add node" screen to say that any non-default disk paths will not be propagated to the node and that they must be set prior to it being added. This helps the most common case of adding a node through the UI, but does not help the CLI or REST API paths.
3 - Expand the "add node" UI interface to allow for applying a non-default disk path when adding the node. This also only covers the UI but not the CLI or REST API.
4 - Change the behavior of ns_server to propagate these disk paths as part of the global configuration. I'm less a fan of this since it changes the existing behavior and may make it harder to have per-node settings here which have been useful in the past.

For #2 and #3, I would argue that improving the UI interface should be the primary goal and that we can cover the difference in the CLI and REST API through documentation on "how to programmatically manage couchbase" (which we do not yet have: http://www.couchbase.com/issues/browse/MB-8105)

Assigning to Anil for triage and prioritization.