I implemented RT ghost
rt_out tasks. They are both implicit tasks (at least for now).
rt_out task is made every time we run with
--radiation, but the
rt_in depends on what other physics are involved. If the feedback is on, I use the
stars_out implicit ghost task instead as the "entry point" to RT. If the
stars_out task isn't present, instead the
rt_in is created and depends on
This way, the RT doesn't depend on running with feedback nor with star formation.