Fix to allow compiling C with Intel CC

[LKedward]

• Minimal set of changes to invoke a separate C compiler instead of passing C code to the Fortran compiler
• Allows compiling packages with C code using Intel compiler collection, Fixes Cannot Compile C code with Intel #426
• C compiler is chosen automatically based on the Fortran compiler selection
  • Companion compiler chosen if known, otherwise fall-back to invoking the Fortran compiler for C as done currently
  • C compiler can alternatively be specified explicitly using environment variable FPM_C_COMPILER

Compiler ID	FC	CC
gcc	gfortran	gcc
intel_classic	ifort	icc
intel_llvm	ifx	icx
intel_flang	flang	clang
ibmxl	xlf	xlc

Not addressed in this PR: specifying C compiler flags

[awvwgk]

Looks good to me 👍.

I wonder if we should aim towards making build_target a method of a more abstract compiler or linker object to separate the vendor/platform specific build logic from our targets in the long-term. This could be useful for the ar vs link issue as well.

[LKedward]

Good point, I agree. That seems like a natural direction to take with the targets in the future.

[ivan-pi]

The Rust cc crate converged to having separate Build (a builder for compilation of a native library) and Tool (configuration used to represent an invocation of a C compiler) abstractions. Their is of course the important difference that fpm is more like cargo and the cc crate all in one, so there is perhaps more potential to share information across layers.

[everythingfunctional]

Should this do something about flags when compiling C code? I'm fine if we want to add that as a separate PR, just curious.

Also, my installation of oneAPI on Windows doesn't seem to have icc. It works as expected for everything else I tried, (gfortran, ifort, and ifx on Windows, Mac and Linux).

Otherwise, I think this looks good.

[LKedward]

Should this do something about flags when compiling C code? I'm fine if we want to add that as a separate PR, just curious.

I thought that best left for a separate PR, since it will require some discussion about best approaches.

[LKedward]

Thanks for the reviews. I'll merge later today if there are no further comments.