Add CJK and Eastern-Slavic name romanization round-trip coverage

[maziyarpanahi]

Summary

Adds a dependency-free romanize_name helper for deterministic Latin renderings of non-Latin PERSON surrogates. It reuses the existing Unicode script detection and segmentation utilities, with algorithmic Kana, Hangul, Cyrillic, and Greek handling plus deterministic Han-name readings and an explicit ASCII fallback for unknown Han characters.

Adds synthetic offline regression coverage for Japanese, Chinese, Korean, Russian, and Greek language packs. The tests exercise detect → replace → re-detect stability, romanized PERSON re-detection, and exact offsets on mixed Latin/non-Latin lines. Unwired language packs skip with a clear wiring reason.

Validation

• .venv/bin/python -m pytest tests/unit/core/test_name_romanization_roundtrip.py -q — 15 passed
• .venv/bin/ruff check openmed/core/translit.py tests/unit/core/test_name_romanization_roundtrip.py — passed
• .venv/bin/ruff format --check openmed/core/translit.py tests/unit/core/test_name_romanization_roundtrip.py — passed
• git diff --check — passed

All test data and detector responses are synthetic and offline.

Closes #468

[maziyarpanahi]

Thank you @maziyarpanahi. I reviewed this against #468 and current master, then added maintainer follow-up d2164502.

What changed:

• Kept the mixed-script name round-trip coverage fully offline and deterministic.
• Exercised transliteration and exact surrogate re-detection without adding network or restricted-data dependencies.

Validation:

• .venv/bin/python -m pytest tests/unit/core/test_name_romanization_roundtrip.py -q — 15 passed.
• .venv/bin/python -m pytest tests/unit -q — 11,593 passed, 89 skipped on the exact aggregate release tree.
• Ruff lint and format checks passed.

This now satisfies the issue and is ready to merge.