Improve Active/Active configuration for the manila share service

Manila's service components are designed to be fault tolerant and resilient. Deployers have the ability to run any number of manila-api, manila-scheduler, manila-share and manila-data processes as necessary with this fault tolerant design. However, there are some theoretical pitfalls and a lack of testing in the upstream community with manila-share running in an active-active configuration.

Pitfalls/concerns:
    - Recovering from service failures
    - Concurrency control for polling operations
    - Monitoring health of an individual host driver service

Use Cases:
    -HA to survive single node failure where pacemaker type control for active-standby support is not available
    - edge deployments with tripleo
    - non tripleo deployments with just haproxy and no pacemaker-corosync

An example of setting up two manila-share services in an active/active HA config:

manila.conf on node 1:
------------------------------
[DEFAULT]
...
host = manila-share-hostname
enabled_share_backends = backend1, backend2
share_topic = manila-share
...
[backend1]
...
[backend2]
...

manila.conf on node 2:
------------------------------
[DEFAULT]
...
host = manila-share-hostname ### Common hostname between
enabled_share_backends = backend1, backend2
share_topic = manila-share
...
[backend1] ### Common backend names in the two services
...
[backend2]

With the above configuration, both services appear to be the same, but there's no unique way to reference one of them, for example, if "manila service-list" is executed, manila-share services corresponding to backend1 will only appear once even though there are two separate services.