RT: remove injection tasks

This MR introduces the following changes:

• move the injection step of RT into the star density and feedback loops. We require a preparation/weights counting loop first, which is now done during the star density loop, and the actual injection step, which is done during the star feedback loop.
• remove all mentions/instances of HYDRO_CONTROLLED_INJECTION. You convinced me to let this one go.
• cleanup of files that aren't used any longer.
• cleanup of task generation/activation for RT
• cleanup and update of debugging checks.

It would be awesome if you could process this MR asap, since its contents most likely will make current issues with the subcycling and the MPI easier to solve.

@tkchan : You will most likely need to check the way you're doing injection again in your current MR, regardless of which MR is merged first. We now need to reset some star variables before the density loop, so some contents of rt_reset_spart() might need to be moved into rt_init_spart(). Also not all variables that used to be available for the injection_prep interactions are now available for it, if they're being computed during the star density loop. (E.g. I used to use spart->density.wcount, which had to be eliminated.)