docs: Fix crash and simplify git calls in mkdocs generation

[petrasovaa]

This is trying to resolve regression after #7310 causing problems in addon compilation in CI. To be honest I am not sure I quite understand the exact mechanism, why gs is None in CI, but this PR should hopefully address the problems in logs like:

Traceback (most recent call last):
  File "/home/runner/install/grass86/utils/mkhtml.py", line 65, in <module>
    get_version_branch(
    ~~~~~~~~~~~~~~~~~~^
        major,
        ^^^^^^
        urlparse.urljoin(base_url, "grass-addons/"),
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ),
    ^
  File "/home/runner/install/grass86/utils/mkdocs.py", line 76, in get_version_branch
    branch = gs.Popen(
             ^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'Popen'

This PR replaces gs.Popen/decode calls with standard Python calls.
One change I am not sure about is in get_addon_path where if gs was not available, it replaced fatal with sys.stderr.write, which is not the same, it does not exit. That seemed like unintentional, so I replaced it with sys.exit, but I am not sure about that.

[petrasovaa]

The CI addon compilation reports success for almost all addons.

[tmszi]

My opinion is that better solution is switching from the Make to CMake (which setting -DPYTHON_EXECUTABLE) for compiling addons in the CI. CI use compile_addons_git.sh BASH script with Make command for addons compilation, but g.extension use CMake.

gs is None because Make command in CI compile_addons_git.sh doesn't initialize PYTHON_PATH ENV.

[echoix]

My opinion is that better solution is switching from the Make to CMake (which setting -DPYTHON_EXECUTABLE) for compiling addons in the CI. CI use compile_addons_git.sh BASH script with Make command for addons compilation, but g.extension use CMake.

gs is None because Make command in CI compile_addons_git.sh doesn't initialize PYTHON_PATH ENV.

It might be a good next step, let’s keep the idea

[neteler]

@petrasovaa to be backported to G8.5?

[petrasovaa]

I am not sure whether 8.5 has the same problem. It's probably harmless to backport.

[neteler]

Backporting worked after cherrypicking 951be1f (#7310)