Fill ib_markers halo cells in post_process

[sbryngelson]

Fixes #2010

post_process writes ib_markers including its ghost layers, but reads only the interior, so the ghosts were 0. The viz silo assembler let those zeros overwrite the neighbor rank's real cells, showing 2 fluid planes inside the body before each internal rank boundary.

Add s_exchange_ib_markers_halo, which copies the markers into a real scalar_field and fills the MPI-boundary ghosts with the existing s_mpi_sendrecv_variables_buffers, then returns the ghost-inclusive array that s_save_data writes. No new exchange code. Physical-boundary ghosts are left as before.

Verified on test A2036630 (2D IBM rectangle, 50x40, CPU+MPI, silo output at step 1), solid cells per x column for columns 15-34:

run	columns 15-22	columns 23-24	columns 25-34
1 rank	16	16	16
2 ranks, master	16	0	16
2 ranks, this PR	16	16	16

The 2-rank output with this PR is identical to the 1-rank output. ./mfc.sh test --no-mpi --no-gpu -j 24 --only IBM passes (61 passed, 0 failed), and ./mfc.sh precheck passes. Only post_process output changes; simulation results are unaffected. The 3D multi-rank case from the issue was not re-run.

---

Acknowledgement

• I confirm this PR meets the above expectations and reflects my own understanding and real-world context.

[github-actions]

Lines of Code

File	Lines	Diff
src/post_process/m_start_up.fpp	808	+16

Directory	Lines	Diff
post_process	3515	+16
total	47442	+16