fix: refresh periodic ghost image shifts with skin

[19hello]

Reminder

• I have read AGENTS.md and docs/developers_guide/agent_governance.md.
• I have linked an issue or explained why this PR does not need one.
• I have added adequate unit tests and/or case tests, or explained why not.
• I have listed the exact verification commands run and their results.
• I have described user-visible behavior changes, including INPUT parameter changes.
• I have explained core-module impact for ESolver, HSolver, ElecState, Hamilt, Operator, Psi, or other source/ changes.
• I have requested any needed governance exception below.

Linked Issue

No linked issue. This PR fixes a correctness issue discovered during local molecular-dynamics boundary-crossing validation.

Unit Tests and/or Case Tests for my changes

• Commands run:

cmake -S /tmp/abacus-md-skin-dynamic \
      -B /tmp/abacus-md-skin-dynamic-build \
      -DBUILD_TESTING=ON \
      -DENABLE_MPI=ON \
      -DENABLE_LCAO=OFF \
      -DENABLE_LIBXC=OFF \
      -DCMAKE_BUILD_TYPE=Debug

cmake --build /tmp/abacus-md-skin-dynamic-build \
      --target MODULE_CELL_NEIGHBOR_mdcell_migrate_mpi \
      -j4

git diff --check

python3 tools/03_code_analysis/agent_governance_check.py --staged

- Result summary:
  - CMake configuration passed.
  - MODULE_CELL_NEIGHBOR_mdcell_migrate_mpi compiled and linked successfully.
  - git diff --check passed.
  - The staged governance check passed with a documentation-sync warning. The warning is addressed below because this change is internal and does not modify INPUT behavior.
  - Added MdCellMigrateMpiTest.SkinUpdatesPeriodicGhostImageWithoutRebuild, which verifies that a periodic ghost image is updated without rebuilding the neighbor-search object.
- Checks not run, with reason:
  - The 2-process MPI test was attempted, but the local runtime environment mixes toolchain and system BLAS/LAPACK/ScaLAPACK libraries. The executable exits before test startup with ABI errors involving gotoblas and sgetrs_N_parallel. This is a local runtime-library issue, not a test assertion failure.

### What's changed?

- Fixed periodic ghost coordinates during reusable-skin MD updates.
- Ghost image coordinates are now recalculated from the current atom position and ghost exchange slot instead of reusing a stale periodic image shift.
- The existing ghost layout and neighbor-search object are reused; crossing a periodic boundary does not force a complete ghost or neighbor-list rebuild.
- The ghost position message now carries both wrapped fractional coordinates and dynamically selected image coordinates.
- Added an MPI regression test covering an atom moving from frac.x = 0.99 to frac.x = 0.01 without neighbor-list rebuilding.
- No user-visible INPUT behavior changed.

### Governance Notes

- INPUT/docs changes: None required. This PR changes only internal MD ghost-coordinate communication and does not change any INPUT parameter, default value, parser behavior, or user-facing documentation.
- Core module impact: The changes are limited to MDCell neighbor decomposition and ghost-coordinate synchronization under source/source_cell/module_neighlist. There are no changes to ESolver, HSolver, ElecState, Hamilt, Operator, Psi, or electronic-structure algorithms.
- Exceptions requested: None.