Fix: Support out_current=2 with multiple MPI processes

[AsTonyshment]

Linked Issue

No linked issue. The bug was reproduced locally.

What's changed?

out_current = 2 was introduced in PR #6886 as a matrix-commutation method for calculating current. Its multi-process path was already known to crash when it was introduced, and this bug has remained unresolved since then.

out_current = 2 incorrectly used global orbital and band indices to access MPI-local distributed matrices, causing heap corruption with multiple MPI ranks. It also did not reduce rank-local current contributions.

This PR:

• uses local indices when constructing the position matrices;
• converts global band indices to the local ScaLAPACK layout;
• reduces current contributions across MPI ranks.

out_current = 2 now works with multiple MPI ranks and remains numerically consistent with the single-rank result. The related parameter and developer documentation was also clarified.

Unit Tests and/or Case Tests for my changes

Case test: the same LCAO RT-TDDFT calculation (bulk Si with 2×2×2 k-point sampling) was run with one MPI rank and 32 OpenMP threads, and with four MPI ranks and 8 OpenMP threads per rank. The currents in current_tot_comm.txt overlap throughout all 10 time steps.

Governance Notes

• INPUT/docs changes: Description only; the accepted values and default are unchanged. The runtime metadata, YAML, and input reference are synchronized.
• Core module impact: Limited to the source_io RT-TDDFT current-output path; no core interfaces or workflows are changed.
• Exceptions requested: None.

[ESROAMER]

Nice work!