Fix to #363. Drifting decision needs to use the same multipole information as the actual tree walk. Also fix some gravity_cache padding values.

their Initial Conditions files. This avoids error rasises from H5G* when
opening files without units and reverting back to the ones specified in
the parameterfile. We now also let users know when their units are being
read from their ICs explicitly.

Applied @jwillis' correction to the second loop of DOPAIR2. dx_max was entering the external loop with the wrong sign.

Modify the inner-loop exit condition to incorporate the distance on the axis calculation rather than pre-compute it.

There is a problem when the cpu mask is empty and using multiple MPI ranks on a node, in that case all the main threads end up pinned to one core during the reading of the initial condition (and any other tasks done before engine_init is called).

In DOPAIR2(), once we have a pj candidate, check the distance on the axis using hj instead of relying only on the earlier hj_max check.

Reverse the loop to find the exit condition in DOPAIR2() to go from the centre to the outside. Follows @jwillis' suggestion in #362

Force compiler to report shadowed variables. Removed all shadowed variables. Added the function scheduler_activate_send() to simplify the unskipping of MPI sends.

Reinstate missing pointer indirection for external gravity dependencies that had disappeared in merge.

DOSELF_SUBSET, DOPAIR1_NAIVE and DOPAIR2_NAIVE should do their calculation in the frame of the cell.