Add explicit regression test for issue #363

[krystophny]

Summary

Closes #363.

Current master already emits the selective use ..., only: import needed to avoid the name collision reported in #363, but there was no dedicated regression coverage for that exact case.

This PR adds an issue363_missing_only_import example that:

• reproduces a module-level variable / procedure-argument name collision
• asserts that the generated wrapper uses use global_mod, only: print_number
• runs through both the regular wrapper flow and the Direct-C Meson flow

Verification

Historical failure on v0.3.0

$ /tmp/f90wrap-v030-venv/bin/python -m f90wrap -m global_mod global_mod.f90
$ sed -n '1,8p' f90wrap_global_mod.f90
! Module global_mod defined in file global_mod.f90

subroutine f90wrap_global_mod__print_number(number)
    use global_mod
    implicit none
...
$ gfortran -c global_mod.f90
$ gfortran -c f90wrap_global_mod.f90
f90wrap_global_mod.f90:7:33:

    7 |     integer, intent(in) :: number
      |                                 1
Error: Name ‘number’ at (1) is an ambiguous reference to ‘number’ from current program unit
f90wrap_global_mod.f90:3:45:

    3 | subroutine f90wrap_global_mod__print_number(number)
      |                                             1~~~~~
Error: Symbol ‘number’ at (1) has no IMPLICIT type

Current branch

$ cd examples/issue363_missing_only_import
$ PATH="/tmp/f90wrap-venv/bin:$PATH" PYTHON=/tmp/f90wrap-venv/bin/python make test
...
Issue #363 test passed: generated wrapper uses selective import

$ PATH="/tmp/f90wrap-venv/bin:$PATH" PYTHON=/tmp/f90wrap-venv/bin/python DIRECTC=yes make -f Makefile.meson test
...
Issue #363 test passed: generated wrapper uses selective import

$ cd /tmp/f90wrap-test-run
$ PATH="/tmp/f90wrap-venv/bin:$PATH" /tmp/f90wrap-venv/bin/python -m unittest test.test_directc
.............................
----------------------------------------------------------------------
Ran 29 tests in 0.014s

OK