Add the Paraview parallel output format (.pvtu with one piece per rank) - #2977

Proposed Changes

PARAVIEW already writes in parallel: all ranks write into one shared .vtu file with MPI-IO. But that file has only one piece, so ParaView reads it on one process, even when ParaView runs with many processes (pvserver or pvbatch with MPI). A large case may not fit in the memory of that process, and reading, filtering and rendering are serial.

This PR adds two output formats that ParaView can also read in parallel:

• PARAVIEW_PARALLEL
• SURFACE_PARAVIEW_PARALLEL

They write a .pvtu file plus a folder with one .vtu piece per MPI rank:

flow.pvtu
flow/flow_0.vtu
flow/flow_1.vtu
...

• ParaView with several processes (pvserver or pvbatch with MPI) gives the pieces to its processes, so each process reads and works on only part of the mesh. It does not need the same number of processes as the SU2 run.
• Each rank writes its own piece, with no collective call, so this is also the fastest format to write.
• Each piece holds the cells of its own partition. Only the points on the partition boundary are written twice: +3% points on 4 ranks and +8% on 16 ranks for a 1.9M-cell mesh.
• The precision follows VOLUME_OUTPUT_PRECISION. The connectivity uses Int64 when a piece needs it.

PARAVIEW is not changed. Use it for small cases and when one file is easier to handle.

Test case

New regression test pvtu_writer / pvtu_surface_writer (parallel, 2 ranks), TestCases/output_writers/pvtu_output.cfg. It writes the flat plate solution with both new formats and compares the .pvtu files with reference files. The pieces are binary, so the test harness cannot compare them.

I checked the pieces by hand on a wing mesh (1.1M cells, 4 ranks), loaded in ParaView 5.13:

• flow.pvtu has the same 1 086 181 cells as flow.vtu from PARAVIEW.
• All 23 fields are identical to flow.vtu (max difference 0).
• surface_flow.pvtu loads correctly. 2 of its 4 pieces are empty because those ranks have no surface elements.
• Unsteady run (NACA0012, 3 time steps, 2 ranks): each saved step gets its own .pvtu file and folder (flow_00000.pvtu, flow_00000/, ...), with the same names as PARAVIEW. ParaView loads flow_*.pvtu as one time series.

Related Work

• Together with Group CGNS boundary sections into markers using the ZoneBC of the file #2976, this PR and that one are the last PRs of the input/output improvements that started with Fix integer overflows and parallelize the output writers for large meshes #2924.
• The reference files are in Reference files for the Paraview parallel writer test TestCases#213.
• Temporary: .github/workflows/regression.yml uses the paraviewParallelOutput branch of TestCases until Reference files for the Paraview parallel writer test TestCases#213 is merged. I will set it back to develop before merging.

PR Checklist

• I am submitting my contribution to the develop branch.
• My contribution generates no new compiler warnings (try with --warnlevel=3 when using meson).
• My contribution is commented and consistent with SU2 style (https://su2code.github.io/docs_v7/Style-Guide/).
• I used the pre-commit hook to prevent dirty commits and used pre-commit run --all to format old commits.
• I have added a test case that demonstrates my contribution, if necessary.
• I have updated appropriate documentation (Tutorials, Docs Page, config_template.cpp), if necessary.

🤖 Generated with Claude Code

https://claude.ai/code/session_01L7oxzVir4zENrpy3K3W6Fm