Call the branching function for the stars interaction instead of the direct SELF/PAIR interaction function.

Construct hydro tasks if there are hydro particles OR star particles in a cell when running with feedback.

When allocating the bisection table for the star smoothing length iteration, use the star particle count, not the gas count.

Correctly activate all the send/recv and drifts of stars. Can now run EAGLE-6 over MPI with everything switched on.

Grackle cosmo

See merge request !740

Can survive the 0th step and up to the first serious rebuild with the current setup and feeback switched on.