Instead of hardcoding 2 runners to be registered, you can add profiles
to a directory to specify what runners should be registered.

Instead of modifying the docker-compose file, it's easier and better to
use a `.env` file where you can provide the parameters for this runner.

In order to be able to run docker images inside docker, we mount the docker
socket inside the build container. This offer a security risk, as any CI job is
able interact with the docker engine the host.

To mitigate this, one runner is used to build docker images. This runner
should be limited to trusted projects.

The other runner can be shared, as it does not get the docker socket.

Instead of having to manually run a command to register the runner,
automatically register the runner on start if the config file does not
exist yet.