bugfix: Avoid PETSc IS size query during boundary rebuild

[gthyagi]

Summary

Fixes #285.

This changes the UW_Boundaries rebuild in _dm_unstack_bcs() to test the returned PETSc IS object by truthiness instead of calling getSize() before setting the stratum.

Root cause

Imported DMPlex meshes can include non-boundary labels such as an Elements / cell-region label. During boundary label reconstruction, label.getStratumIS() may return an empty or otherwise problematic PETSc IS for those labels. Calling label_is.getSize() on that object can trigger a PETSc-level abort before the Stokes system is assembled, which is what issue #285 reports for tests/test_1012_stokesImportedDMPlex.py::test_stokes_essential_bc_imported_dmplex.

The new check follows the existing safer PETSc-object truthiness pattern and avoids dereferencing the IS only to determine whether it is usable.

Validation

• ./uw build
• PYTHONFAULTHANDLER=1 ./uw python -X faulthandler -m pytest tests/test_1012_stokesImportedDMPlex.py::test_stokes_essential_bc_imported_dmplex -ra -vv -s passed
• ./uw python -m pytest tests/test_1012_stokesImportedDMPlex.py tests/test_0502_boundary_integrals.py -ra -q passed: 23 passed
• ./uw python -m pytest tests/test_0001_meshes.py tests/test_0830_mesh_adapt_variable_transfer.py -ra -q passed: 16 passed, 5 skipped
• ./uw python -m pytest tests/test_1001_poisson_constants.py -ra -q passed: 6 passed
• ./uw test completed with unrelated existing failures in memprobe/kdtree and AMR swarm migration tests: 6 failed, 618 passed, 14 skipped, 753 deselected, 7 xfailed, 1 xpassed

Notes

The patch is intentionally minimal: it preserves _dm_unstack_bcs() label reconstruction behavior while avoiding the PETSc IS.getSize() call that reproduces the segfault path.