Testing of CLI clients

Motivation:
Tempest is quite poor for CLI tests now and this should be general proposal of idea how to do it..

Tests will compare CLI outputs with outputs from python client library call.

It should covers components and switches as much as possible. In example:
* glance
    image-create
    image-list
    ...
* nova
    image-list
    network-list
    hypervisor-list
    boot ...
* keystone (update/extend current files)
* ... (other components with CLI interface)

Step by step workaround:
Startup: connect to remote endpoint by SSH (via Paramiko i.e.)
1. execute particular command remotely
2. catch output 'as it is' (like "glance image-create") + processing (parsing
3. match output (2.) with result given by python client library
Teardown: close SSH connection, cleanup (if any)