stdcm: test that an extra fake track actually works at faking stops

[eckter]

See https://github.com/osrd-project/osrd-dags/issues/675

Before tweaking with the data, we need to see if it actually works.

We'd need to:

Create an infra with one long line and several signals, and one side track in the middle:

          ___
         /    \ 
________/______\_________

The tracks before and after need to be very long with several signals every 2km.

We then add signals and detectors: one at the start of the side-track, then 3 at the end, 1m apart.

We create a timetable where lots of close trains on the main track, too close for a new train. One opening on the first half and one opening on the second, but far apart, so we need to stop.

Then, we need to find a solution where it goes on the side track.

[eckter]

After running some tests on the infra added in #10657:

This generally works. But there's several conditions:

1. We need a signal to be placed at least $convoy_length from the start of the track. There may be several of them for convoys of different length.
2. We only need two signals at the end
3. We must not see signals from the main tracks in the engineering allowance section, so the last signals must be 400m away from the end of the takeover track
4. The engineering allowance itself must be contained in the takeover track, we can't add slowdown parts before. But there will be a slowdown automatically added to match the max speed of the takeover track => it needs to be low enough. 30km/h seems fine but we could lower it if needed.

And then there's newly discovered bugs to fix.