[LIBSTELL] Use a more direct (i.e. no loops) implementation of mpi_calc_myrange

[jonathanschilling]

This is a further iteration of #240.

This is a fundamental and substantial change to a piece of code used by many routines.

I agree and therefore, I would say that it would be valuable to have this routine very well documented and almost obvious to understand.
In the suggested change, the assumptions and the approach are (I hope) quite explicitly documented and reasoned about. Please let me know if you think differently about this.

I also attached a stand-alone test program (test_mpi_calc_myrange) now that I used to verify that the computed mystart and myend are identical (for the probed parameter value ranges) to what the old implementation produced:
test_mpi_calc_myrange.zip
It is compiled with gfortran test_mpi_calc_myrange.f -o test_mpi_calc_myrange
and then run with ./test_mpi_calc_myrange and should then print all tests passed successfully.
I understand that it was not obvious from the previous iteration of this PR that the outputs are identical. Sorry for that.

[lazersos]

@jonathanschilling This looks better. All your types are integers and so integer math is used, this could introduce issues. I'll double check this today and get back to this.

[jonathanschilling]

Thanks! The use of integer math is intended, as the remainder of the division is explicitly handled via the work_remainder variable. Do you know a way to make it more explicit in the code that integer division is intended? (I assume that you are concerned about the division in work_per_rank = total_work / local_size, right?)

EDIT: added comments in the code to preserve this discussion