glusterfs_native breaks multibackend

[Csaba Henk]

steps of reproduction:

- create a multibackend setup so that
    - one of the backends is glusterfs_native
    - there is another backend B
- create a share for B
- stop m-shr service
- try to restart m-shr: it will deterministically crash

Note: the issue was found by code analysis by Igor Malinovksy and further investigated by Csaba Henk; however, we haven't seen the issue in situ,
so there is no backtrace or error report available.

[Csaba Henk (chenk)]

Note that following change: https://review.openstack.org/174418 fixes it in master (although was submitted for another bug).
The Kilo backport request is to be understood as a request for backporting this change.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/183435
Committed: https://git.openstack.org/cgit/openstack/manila/commit/?id=49e223e2e67797fb8887b0f7052b5f3c6969aa06
Submitter: Jenkins
Branch: stable/kilo

commit 49e223e2e67797fb8887b0f7052b5f3c6969aa06
Author: Igor Malinovskiy <email address hidden>
Date: Mon Apr 20 18:16:07 2015 +0300

    Remove direct DB calls from glusterfs_native driver

    - Remove direct DB calls from methods create_snapshot() and
    delete_snapshot().
    - Move code from _update_gluster_vols_dict() method to
    ensure_share() method.

    Note that the original of this change was submitted as
    a fix to #1444914; however, we've backported it to
    Kilo for the purpose of fixing an other bug, as indicated
    below.

    Change-Id: I6c95f6d9361093d832a536971a460c3cdda44dcb
    Closes-Bug: #1447692
    (cherry picked from commit 4eb57f53c1c705fa3abf2ab0c2973fe709dbe6c8)