Passive swarm save/advection hangs or crashes on empty MPI ranks

[bknight1]

Problem

Passive swarm operations are not safe when one or more MPI ranks hold zero particles. Observed symptoms:

• hangs during write_timestep / swarm save — the core reported bug
• hangs in evaluate / global_evaluate when evaluating swarm particle data
• estimate_dt() crashing on an empty rank with ValueError: cannot reshape array of size 0 into shape (0,newaxis)

What we found (after rebasing onto development)

The evaluate / global_evaluate / advection empty-rank deadlock was already fixed upstream by #611 / PR #656 (forward the cell-location policy on the empty-rank path). With that merged we verified empirically (np=4) that the default non-evalf path now works on empty ranks, so the earlier evalf-threading workaround was dropped.

The remaining, actual cause of the write_timestep / save hang is a parallel-HDF5 (MPIO) collective-close deadlock triggered by a per-rank dtype mismatch:

• An int swarm variable (e.g. swarm.add_variable("uid", size=1, dtype=int)) holds particles on some ranks but not others.
• On an empty rank, SwarmVariable.unpack_raw_data_from_petsc() returned np.zeros((0, n)) — a float64 array — losing the field dtype, so an empty rank reported float64 while a non-empty rank reported int32.
• SwarmVariable.save() passes the per-rank local_data.dtype to the collective parallel-HDF5 create_dataset(...). Divergent dtypes leave HDF5 metadata inconsistent, so the collective close never synchronises and the save deadlocks (same silent-hang class as [BUG] - swarm.write_timestep fails in parallel (np > 1) #151).

Fix

1. estimate_dt() — empty-rank guard: zero-particle velocity data is reshaped to an explicit (0, dim) array instead of crashing on reshape(n, -1).
2. Empty-rank dtype preservation — store the field PETSc dtype (_petsc_dtype) and use it in every zero-length fallback array, so empty-rank data agrees with the field true type across ranks. Fixes the MPIO save deadlock.

Tests (passing at np=2 and np=4)

• tests/parallel/test_0795_swarm_empty_rank_evaluate_save.py — evaluate + write_timestep on empty ranks. test_passive_swarm_save_empty_ranks reproduces the user pattern and hung at np=4 before the dtype fix.
• tests/parallel/test_0796_swarm_advection_empty_rank.py — advection on the default path with empty ranks (guards the test_global_evaluate_after_migration hangs at np=4, and its file is in neither parallel-test glob #611 regression).

Both files: 5 passed at np=2 and 5 passed at np=4.

Related empty-rank / global-evaluate work: #611.