walberla: fix LB VTK writers exporting last block's data on multi-block ranks (bug-sweep #14)

[RudolfWeeber]

register_vtk_field_writers installed one VTKWriter instance plus a
beforeFunction that looped over all local blocks, repeatedly calling
set_content/set_dims on a single shared buffer. waLBerla runs the
beforeFunction exactly once, so the buffer retained only the LAST
block's data; the subsequent per-block evaluate pass then wrote that
same buffer for every block. On any MPI rank holding more than one
waLBerla block (blocks_per_mpi_rank with >1 block), density,
velocity_vector and pressure_tensor VTK output was corrupted for all
blocks but the last. Single-block ranks were unaffected, masking the
bug. BoundaryVTKWriter already reads per-block via configure()/block_.

Fix: make the VTKWriter buffers block-aware. m_content/m_dims become
per-block maps keyed by the IBlock pointer, filled by the beforeFunction
for every local block. configure() (invoked per block with block_ set)
selects the current block's buffer, so evaluate()/get_first_index read
the correct per-block data. The single-block path is numerically
identical.

Adds LBWalberlaVTKDoublePrecisionBlocksCPU to lattice_vtk.py
(blocks_per_mpi_rank=[2,1,1]) which fails on the unfixed code with a
~50% per-node velocity/density mismatch and passes after the fix.

Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com

🤖 Generated with Claude Code

[espresso-ci]

Your pull request does not meet our code documentation rules. The Doxygen documentation generated 1 unique warnings (total: 1, ignored: 0):

src/lattice_boltzmann/LBVTK.impl.hpp:
  line 62: unable to resolve reference to 'block_' for \ref command

You can generate these warnings with make -t; make doxygen; ../maintainer/CI/dox_warnings.sh using the maxset config. This is the same command that I have executed to generate the log above.