MultiPath creates non-specific path_id

[jnspitale]

When '+' is supplied as the path_id, MultiPath registers the supplied paths using the ID of the first path with a '+' appended. This path_id could describe any number of possible collections of paths, and caused a conflict in the metadata code. The immediate solution was to not cache the paths. That solution is fine, but the conflict took time to track down, and could do so again in the future if not fixed.

My proposed solution, implemented in ideas/multipath/multipath.py, is for the path_id to be comprised of the names of all of the included paths in alphabetical order. Alphabetization is not necessary, but it helps to ensure that an identical group of paths are not cached more than once if they are specified in different orders.