CI/CD proposition for xdp and qp

[sebastian-mutz]

This was discussed around PR #1133 and proposed to tackle in a separate PR. I'm creating an issue to keep things organised/visible (and to keep track of it after the current PR is merged).

Background

Building with flags FFLAGS="-O3 -DWITH_QP=False -DWITH_XDP=False" failed after PR #1099 was merged (see issue #1131), but the tests should ideally have caught this before.

Discussion (#1133)

@jalvesz:

[...] we are not properly testing this configuration which should have been tested. Would it be possible to check if we could adapt one of the CIs (either with CMake or fpm) to check for the case of -DWITH_QP=False -DWITH_XDP=False

[...]

Currently, the CMake builds enable by default qp ( and xdp with gfortran) as it checks if available in the compiler and then allows it. Since we only check with gfortran, ifort and ifx which have qp and xdp for the first. It all passed.

With FPM, we added a job to test for xdp and qp. Currently, the FPM jobs are slower than the CMake jobs. I would like to propose then to use that job to test "without" then. In this case, we should first make sure that it does fail with fpm when desactivating both kinds and without the current fix.

https://github.com/fortran-lang/stdlib/blob/ee330556a02987d006f1089c44837d73505e4436/.github/workflows/fpm-deployment.yml#L45C1-L47C67

The last point was addressed in PR113 (commit 4c38baa)

@jvdp1:

Re: CI/CD for xdp and qp, I support @jalvesz 's proposition. An otpion is to add another pipeline (in another PR) specifically for that and similar to the ones used for the modularization of stdlib.