Changes

ttcn3-s1gw-test: run eNB/MME domains in separate networks

The idea is to simulate more realistic conditions for the IUT by
creating two virtual network interfaces in each container:
one for the eNB (A) and one for the MME (B) domains.

                (network A) | (network B)
                            |
                     +-------------+
                     |  osmo-s1gw  |
                     +-------------+
  +----------------> |             | <---------------+
  |                  |             |                 |
  | +--------------> |             | <-------------+ |
  | |                +-------------+               | |
  | |                       |                      | |
==|=|=======================|======================|=|==
  | |                       |     +-----------+    | |
  | |   +-------------+     |     |           |    | |
  | +-> | ConnHdlr(1) | <-------> |           | <--| |
  |     +-------------+     |     |           |      |
  |                         |     |  S1APSRV  |      |
  |     +-------------+     |     |           |      |
  |---> | ConnHdlr(N) | <-------> |           | <----+
        +-------------+     |     |           |
                            |     +-----------+
                            |

This is achieved by passing multiple --network arguments to the
docker-run command.  Sadly, we cannot use docker_network_params
more than once, because docker does not understand multiple
--ip / --ip6 arguments (using the most recent one).

Furthermore, we cannot use network_replace_subnet_in_configs
because it replaces subnet in all IP addresses, so we put %VAR%s
to the config files and replace them using sed manually.

Change-Id: I4c72cb58eeb2d0ccca307dcfcfb04d09bfb0834e
Related: SYS#6772