Update install.sh to use FC=gfortran, CC=gcc, MPI_HOME=/path/to/mpi/if/specified/or/installed

[zbeekman]

I think it might be wise to update the install.sh scripts to pass FC=gfortran, CC=gcc rather than mpif90 and mpicc etc. The most complicated part about this is that I need to do some testing and investigation of how to handle the logic surrounding MPI. I think that if install.sh either:

1. Installs mpich as part of the build, or
2. Receives a user flag setting the MPI location

Then it should automagically "just do the right thing" ™️ IF we also set MPI_HOME environment variable. If the script doesn't do the install of MPICH, and the user (or script elsewhere) doesn't explicitly specify an MPI to use then CMakes FindMPI should pick it up from either:

1. The users path via it's searching for compiler wrappers and mpirun etc.
2. environment modules (gnu modules) setting the MPI_HOME variable explicitly
3. other crazy and complicated magic from FindMPI. When CMake fails on this step it also outputs info so that the user can specify via the command line or ccmake or cmake-gui interfaces the explicit location of where to find MPI.

I think it should be fairly straightforward to implement this, but I need to dig into the install script to assert this with any confidence first....

[zbeekman]

@rouson Currently I'm having trouble following variable scoping... I have a WIP patch that I think will work for build_opencoarrays.sh, but I can't figure out the correct way to get FC and CC into build_opencoarrays.sh. Lets talk about this, or perhaps you can do it for me or tell me how to do it?

Here is my idea (WIP, fails to build/test): https://github.com/sourceryinstitute/opencoarrays/compare/topic-%23269-FC-gfortran but currently complains due to lack of FC and CC.... I.E. they must be explicitly set if the user passes them on the command line via -f and -c, or they must point to the gcc that was installed as part of the build.

n.b. CC could in theory point to a different C compiler