Minor updates to GridGenerator::hyper_cube_with_cylindrical_hole().

[bangerth]

No description provided.

[bangerth]

I made the function more robust. As I discovered in #15689, the function breaks there because we now have different vertex numbers. This specific function relies on specific vertex numbering produced by creating a triangulation by another function. This patch changes things in such a way that we ask about geometric attributes of vertex locations, rather than their numbers.

[kronbichler]

There are a few tests that complain, apparently because they expected a different order of the output:

	1840 - grid/grid_generator_04.debug (Failed)
	1935 - grid/grid_output_input.debug (Failed)
	1992 - grid/manifold_ids_on_boundary_01.debug (Failed)

[bangerth]

The one failing check is the same old petsc_step-27 test that keeps failing elsewhere. We'll have to figure out what is happening with this test at some point...

[bangerth]

Ping?

[marcfehling]

Suggested change

	for (const unsigned int f : GeometryInfo<dim>::face_indices())
	for (const auto f : cell->face_indices())

You moved away from GeometryInfo above. We can also do it here.

[bangerth]

True. It really doesn't matter here because the function only deals with hexes. But as a general rule, it's good to move away from GeometryInfo. So committed, patch pushed.

[kronbichler]

Failing MPI test trilinos/solver_control_06 is unrelated.