Re-organised the creation of the RT-specific time-step tasks on foreign cells to avoid duplicates

Implements a cleaner version of #845 (closed).

This fixes the problem reported in !1595 (merged).

Probably needs some careful testing. It works smoothly on the EAGLE-12 RT example but I have not gone beyond that.

added MPI bug radiative transfer labels

requested review from @mivkov

mentioned in merge request !1595 (merged)

Not sure this needs more testing. It's a critical task. If something went wrong there, we'd be hanging somewhere almost immediately, and sub-cycles wouldn't run.

Agreed that this is likely a good enough stress-test.

merged

mentioned in commit 705c38cb

mentioned in issue #845 (closed)