Generic driver no-share-server setup fails on the CI

[Goutham Pacha Ravi]

Description
===========
The generic driver can be setup in both driver modes: a mode that creates and manages share servers for multi tenancy or one where share servers are pre-configured for use by the driver.

The job "manila-tempest-dsvm-generic-no-share-servers" is a legacy zuulv3 test job that tests the scenario where pre-provisined share servers are configured to be used in the deployment. This job has been failing for a while now because these pre-provisioned share servers are not reachable during either Devstack-gate post test hook, or when the manila share manager tries to start up the driver.

See the build history and logs here:

 https://zuul.opendev.org/t/openstack/builds?job_name=manila-tempest-dsvm-generic-no-share-servers#

Steps to reproduce
==================

The manila devstack plugin can create share servers and configure them in manila.conf for specified backends: https://opendev.org/openstack/manila/src/commit/2277c20c02641c3a0cd0804ef53a54cd5492ce2d/devstack/plugin.sh#L340-L401

This setup can be followed to create a couple of VMs to act as share servers from the manila service image published at: https://tarballs.opendev.org/openstack/manila-image-elements/images/

Expected result
===============
VMs are accessible and share creation succeeds

Actual result
=============
Cannot ping these VMs from the DevStack host; this script fails: https://opendev.org/openstack/manila/src/commit/2277c20c02641c3a0cd0804ef53a54cd5492ce2d/contrib/ci/common.sh#L10-L31

Environment
===========
1. This bug occurs on all versions of manila

2. Which storage backend did you use?
   Generic, multi-backend (2 backends) running in no-share-servers mode.

3. Which networking type did you use?
   Neutron with OpenVSwitch

[Goutham Pacha Ravi (gouthamr)]

This bug was discussed at a community IRC meeting: http://eavesdrop.openstack.org/meetings/manila/2020/manila.2020-05-14-15.00.log.html#l-136

A case was made to not continue running this test job; we will lose test coverage for DHSS=False mode operation of the Generic driver; however, this is a reference architecture, and not expected to be used in production. If anyone feels like this should be tested, we can revive this job and fix it up.