This pull request contains three changes: the first two are fairly trivial and the last one (994c952) fixes an issue with setting the mask. It is quite invasive though is it has to change all scheduling for almost all functions.
AHFinderDirect: Fix use of mask_is_noshrink
If the horizon is allowed to shrink then we do expect that interpolating
will eventually use grid points from within the horizon when looking for
the new horizon, so we do not abort in that case.
AHFinderDirect: Added warning in parameter checking
Currently setting find_every_individual to unequal
frequencies is incompatible with moving origins.
Fix is non-trivial without breaking other features,
so a warning was inserted for the time being.
AHFinderDirect: change scheduling from global to global-early
schedule maybe_do_mask local instead of global loop-local so that it is run
at the correct time for find_every != 1 to work. This makes it necessary to
schedule find_horizons and all other global routines global-early since they
would be called after the local routines otherwise.