feat: Allow using Ray as a drop-in distributed multiprocessing Pool

[adigitoleo]

• Ray has a multiprocessing.Pool API that is very similar to the already supported Python stdlib multiprocessing stuff
• Refraining from a nested imap inside the M-index calculation routines cuts down the memory even more, at the cost of extra CPU time (which can be gained back by running on more nodes now)
• At the moment only the 3d simple shear test is properly hooked up to Ray with the ray.put and ray.get calls, but doing the same for other slow tests would look the same

[adigitoleo]

Example:

#!/bin/bash
#PBS -P xd2
#PBS -q normalbw
#PBS -l walltime=05:00:00
#PBS -l ncpus=168
#PBS -l mem=1536GB
#PBS -l jobfs=2400GB
#PBS -l storage=scratch/xd2
#PBS -l wd
#PBS -o simple3d_unsteady_prescr_dc1000000.log
#PBS -e simple3d_unsteady_prescr_dc1000000.err
#PBS -N simple3d_unsteady_prescr_dc1000000

# Install pydrex into this python environment.
module purge
module load python3/3.11.7 python3-as-python
ulimit -c unlimited

# Copy the tools/pbs_start_ray*.sh into $PWD.
source pbs_start_ray_cluster.sh

# Run this script from outside the PyDRex tree, which should be in `PyDRex/*`.
ray job submit --working-dir PyDRex/ -- \
    python3 -m pytest tests/test_simple_shear_3d.py \
    --outdir=$PBS_O_WORKDIR/_out_1 -v --runslow --ncpus=$PBS_NCPUS -k="direction_change[1]"

ray stop