Error stop improvements

[scivision]

I originally authored this error_stop routine. The logic selecting it in CMake was no
longer relevant. I made a driver for the error stop test to make it more reliable.

oneAPI 2021.1 on MacOS has bugs breaking stdlib_error, so I update CI to latest oneAPI for MacOS.

[jvdp1]

Thank you. LGTM.

[awvwgk]

I think there is some value in keeping the pre-Fortran 2018 error stop workaround. The current default branch still compiles with Intel 19, with this patch we would drop support for all Intel versions before 2021.3.

❯ ifort -V
Intel(R) Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 19.0.5.281 Build 20190815
Copyright (C) 1985-2019 Intel Corporation.  All rights reserved.

❯ ninja -C _build_i19/ -j1
ninja: Entering directory `_build_i19/'
[1/184] Building Fortran object src/CMakeFiles/fortran_stdlib.dir/stdlib_error.f90.o
FAILED: src/CMakeFiles/fortran_stdlib.dir/stdlib_error.f90.o src/mod_files/stdlib_error.mod 
/opt/intel/2019/compilers_and_libraries_2019.5.281/linux/bin/intel64/ifort  -I../src -Isrc/mod_files -module src/mod_files/ -warn declarations,general,usage,interfaces,unused -stand f18 -c src/CMakeFiles/fortran_stdlib.dir/stdlib_error.f90-pp.f90 -o src/CMakeFiles/fortran_stdlib.dir/stdlib_error.f90.o
../src/stdlib_error.f90(39): error #6345: Either a PAUSE, STOP or ERROR STOP statement has an invalid argument.   [MSG]
    if(.not.present(code)) error stop msg
--------------------------------------^
../src/stdlib_error.f90(42): error #6345: Either a PAUSE, STOP or ERROR STOP statement has an invalid argument.   [CODE]
    error stop code
---------------^
compilation aborted for src/CMakeFiles/fortran_stdlib.dir/stdlib_error.f90-pp.f90 (code 1)
ninja: build stopped: subcommand failed.

Also, while Intel 2021.1 compiles stdlib, the support for Fortran 2018 error stop is buggy and will always report exit status 128 even if zero is passed as exit code.

[milancurcic]

@scivision can you explain in a sentence or two what this PR changes or improves? I understand the part about the more robust test driver. What is the tradeoff for dropping support for earlier Intel compilers?