- avoid variable shadowing by adding scope
 - use std::rand()
 - remove implicit and explicit casts
 - rely on actual size of geometry object (polylines)

The test would fail sporadically (say in one out of 10000) runs.
Reasons are not clear. My best guess is that the number of actually
stored polylines differ from the number of generated polylines.

A failing testrun would look like this:

 /Users/gitlab/builds/TqLuBEs-/0/ogs/ogs/Tests/GeoLib/TestDuplicateGeometry.cpp:92: Failure
Expected equality of these values:
  n_plys
    Which is: 6
  new_plys->size()
    Which is: 5
[  FAILED  ] GeoLib.DuplicateGeometry (1 ms)

[App] Partmesh; Exclude integration point data.

Closes #3013

See merge request ogs/ogs!3043

inside the partmesh tool. Keep them otherwise.

Metis converted input prepared with
partmesh -i cube_1x1x1_hex_8.vtu --ogs2metis
mpmetis -gtype=nodal cube_1x1x1_hex_8.mesh 2

For now exclude the partitioning of this mesh item type.

If component transport is not enabled in
the configuration, the variable is unused.

[Doxygen] Fixed project file links

See merge request ogs/ogs!3047

Small fixes

See merge request ogs/ogs!3046

[ci] Build openmpi container.

See merge request ogs/ogs!3044

Update to newest MGIS

See merge request ogs/ogs!3028

Don't modify the behaviour data view.

OGS specific commits applied on top.

[CL] Chemical calculation shifted from nodes to int_pts (Part III): Calling intepolator and extrapolator in Timeloop

See merge request ogs/ogs!3019

[PL/RM] Remove dry density pellet output.

See merge request ogs/ogs!3042