Integration tests

[fmauch]

This aims at fixing #9

The tests in this PR work fine in principle. However, the paremetrization with each robot model seems to lead to a situation where the simulation isn't properly torn down, so for the second tested model there already is a simulation running which prevents the respective simulation from coming up correctly.

[fmauch]

This should now run "successfully".

There seem to be issues using gz sim inside launch_testing. As far as I understand things, gz sim is started in a sub-shell. Then, when the test is stopped, child processes are not necessarily killed, correctly (see ros2/launch#545).

When running our sim_control launchfile, the following processes are started:

auch     478001  9.1  0.1 213548 44864 pts/5    Sl+  14:20   0:00 /usr/bin/python3 /opt/ros/rolling/bin/ros2 launch ur_simulation_gz ur_sim_control.launch.p
mauch     478012  0.0  0.0   2880  1920 pts/5    S+   14:20   0:00 /bin/sh -c ruby $(which gz) sim  -r -v 4 empty.sdf --force-version 8
mauch     478014  2.7  0.1 247956 64176 pts/5    S+   14:20   0:00 gz sim -r -v 4 empty.sdf
mauch     478050 25.5  0.5 3256864 169584 pts/5  Rl   14:20   0:01 gz sim server
mauch     478051 53.3  1.3 4760344 428392 pts/5  Sl   14:20   0:02 gz sim GUI

• Stopping 478012 with SIGINT doesn't seem to do anything.
• Stopping it with SIGTERM keeps the three gz sim processes running.

So, with this, a gz sim instance keeps running after the test finished. I would rather not merge it this way, but I would also rather much like to have integration tests...

[VinDp]

Looks good to me!