[LIE] Fix a bug in LIE#HM when matrix is disabled for the flow process

Closes #3495

See merge request ogs/ogs!5078

based on the GREAT cell experiment simulation

More `mpirun --bind-to none`

See merge request ogs/ogs!5080

This reverts commit c7fc9b4f.

Was set on frontend only.

[web] Fix links

See merge request ogs/ogs!5079

Explicitly "free" 0-core binding of mpirun

See merge request ogs/ogs!5077

Binding to core 0 is default behaviour of mpirun(1):

Please note that mpirun automatically binds processes as of the start of the v1.8 series. Three binding patterns are used in the absence of any further directives:

Bind to core: when the number of processes is <= 2
Bind to socket: when the number of processes is > 2
Bind to none: when oversubscribed

[py] Detect runtime venv in embedded interpreter

See merge request ogs/ogs!5075

Removed hard-coded build-time venv.

[ci] Do cdash submit on failed ctest runs as well

See merge request ogs/ogs!5074

This can happen if you have the ogs-wheel installed but are running a local ogs binary. In this case use `import OpenGeoSys` or uninstall the wheel.

Parallel assembly in HM

See merge request ogs/ogs!5054

Unfortunately with gcc-13.2.1 with optimizations there is a warning, which I was not able to correct.

    ...
    inlined from ‘std::vector<std::tuple<long int, std::reference_wrapper<ProcessLib::LocalAssemblerInterface> > > {anonymous}::collectActiveLocalAssemblers(const BaseLib::PolymorphicRandomAccessContainerView<ProcessLib::LocalAssemblerInterface>&, const std::vector<long unsigned int>&)’ at /var/lib/gitlab-runner/builds/F1XUyv4cx/1/ogs/ogs/ProcessLib/Assembly/ParallelVectorMatrixAssembler.cpp:132:62:
/usr/include/c++/13.2.1/bits/stl_iterator_base_funcs.h:175:9: warning: iteration 576460752303423487 invokes undefined behavior [-Waggressive-loop-optimizations]
  175 |         while (__n--)
      |         ^~~~~
/usr/include/c++/13.2.1/bits/stl_iterator_base_funcs.h:175:9: note: within this loop

Other compilers gcc-14, clang-17 and 18, msvc had no problems with the
former code.

... scheme. Passing a vector of vectors (outer for processes,
inner over the process variables) of residuum names for each
process variable.

This is mostly copy from the VectorMatrixAssembler.

A little bit :) 0.5% maybe. Measured on TRM/Mockup ctest.

This costs some memory (now element ids and local assembler references are stored)
but simplifies the code. If local assemblers would return an element id,
the code could be simplified further.

This slightly simplifies code avoiding extra boolean variable.

This is not sufficient for the staggered scheme yet.