[Fortran] Use mpi_f08 module to fix Intel compiler warnings

[SFrijters]

Description

Intel compiler version 19.1.0.20200306 (and above, presumably) complains:

error #8889: Explicit declaration of the EXTERNAL attribute is required

This can be solved by using the newer mpi_f08 module and accompanying newer datatypes.

Open question: if this change is acceptable, should https://github.com/libocca/occa/blob/main/scripts/compiler/fortranCompilerSupportsMPI.f90 also use mpi_f08?

Alternative: add -warn noexternal to compiler flags to suppress the warning (this was the 'fix' I had in place locally before trying to actually fix it).

[codecov]

Codecov Report

Merging #539 (9a61ea6) into main (88e30e0) will decrease coverage by 0.03%.
The diff coverage is 0.00%.

❗ Current head 9a61ea6 differs from pull request most recent head 9501c78. Consider uploading reports for the commit 9501c78 to get more accurate results

@@            Coverage Diff             @@
##             main     #539      +/-   ##
==========================================
- Coverage   76.59%   76.56%   -0.04%     
==========================================
  Files         263      263              
  Lines       19472    19480       +8     
==========================================
  Hits        14915    14915              
- Misses       4557     4565       +8     

Impacted Files	Coverage Δ
src/c/base.cpp	85.22% <0.00%> (-1.99%)	⬇️
src/core/device.cpp	90.40% <0.00%> (-2.23%)	⬇️

[kris-rowe]

Looks good.

Open question: if this change is acceptable, should https://github.com/libocca/occa/blob/main/scripts/compiler/fortranCompilerSupportsMPI.f90 also use mpi_f08?

Using the mpi_f08 module is safer anyway since it uses explicit interfaces, allowing for type checking. One possible caution against blanket changing this could be if someone is using an older version of OpenMPI or MPICH that doesn't include the mpi_f08 module.

It would probably be safest to have a second fortranCompilerSupportsMPIF08.f90 file for shellTools.sh to use. That being said, after a quick grep through the code it doesn't appear that fCompilerSupportsMPI in shellTools.sh is called anywhere currently— so maybe that can go on the backburner for now.

[SFrijters]

That pretty much follows my thought process exactly.
Do you know what the intended use of fCompilerSupportsMPI is? I mean, apart from the obvious blanket 'checking if the fortran compiler supports MPI'.

[kris-rowe]

Most of the functions in shellTools.sh are used by the Makefile build system.