[Bug] Error in transfer halo particles

[bodhinandach]

Describe the PR
This PR fixes potential deadlocks and crashes in the MPI halo particle transfer routine. The changes are minimal and preserve the original communication pattern. Key fixes include:

• Using stable storage (std::vector<unsigned> send_counts) for particle counts in nonblocking sends, avoiding dangling pointers.
• Correctly sizing send_requests with resize() instead of reserve(), preventing out-of-bounds writes.
• Matching MPI_Probe with MPI_Recv using the probed status.MPI_SOURCE and status.MPI_TAG, eliminating message size/source mismatches.
• Ensuring type consistency during unpack (ptype, nmaterials, mat_id now use unsigned with MPI_UNSIGNED).

These changes eliminate common sources of bottlenecks and undefined behavior, which particularly emerge in large numbers of particle simulations.

Related Issues/PRs
N/A

Additional context
N/A

[codecov]

Codecov Report

❌ Patch coverage is 0% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.73%. Comparing base (e512900) to head (809ba59).
⚠️ Report is 12 commits behind head on master.

Files with missing lines	Patch %	Lines
include/mesh/mesh.tcc	0.00%	27 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #111      +/-   ##
==========================================
- Coverage   95.73%   95.73%   -0.00%     
==========================================
  Files         244      244              
  Lines       51462    51464       +2     
==========================================
  Hits        49267    49267              
- Misses       2195     2197       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[thomaszyu]

Looks good to me!

[bodhinandach]

@thomaszyu, @cgeudeker, and @Curiim, when you have time, would you be able to pull/cherry-pick this branch/commit temporarily to your favourite branch and try doing the usual MPI compile and run? This should not change anything in your simulation.

[thomaszyu]

Hi Nanda, I cherrypicked your bug fix into my force output hardcode branch and ran MPM with 16 MPI ranks, the results look good to me.

[bodhinandach]

@thomaszyu Thanks for checking. I notice there is still a bug... especially when running in full-debug. Will try to understand this further before merging this branch.