Fix MPI with `-g` option
We need to figure out what is going on when using Intel-MPI, vectorization and SWIFT's '-g' option.
@jwillis I'll assign this one to you. Good for you to look at many parts of the code.
Some possible things to look at:
- Run with
- Check that the Sodshock and sedov blasts give the right answer with both MPI libraries and with
- Check whether we have any race condition. Maybe by running with one single thread per rank.
- Check whether affinity pinning helps (check how to do it with Intel-MPI).
Also, it would be good to list the setups that don't work here. Compiler choice + flags + MPI library + SWIFT command line options.