Add conservative Indic morphology boundary refinement

[maziyarpanahi]

Description

Adds an opt-in, precision-first morphology layer for Hindi and Marathi Devanagari plus Tamil, Telugu, Malayalam, and Kannada name spans. It conservatively strips frequent case-marker and postposition suffixes only when both the upstream confidence threshold and an exact caller-provided stem allow-list pass, preserves grapheme-aligned source offsets, and reverses supported Tamil glide sandhi with explicit source mappings. The existing span path remains unchanged unless the layer is enabled.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Code refactoring
• Performance improvement
• Test addition/improvement

Changes Made

• Added conservative cross-script suffix tables, stemming, sandhi splitting, idempotence, grapheme segmentation, and output-to-source offset maps.
• Added an opt-in decoder boundary-refinement hook restricted to canonical person-name labels and guarded by confidence plus an exact stem allow-list.
• Added a synthetic six-language fixture and held-out suffix-like names covering precision, offset reconstruction, over-stripping, idempotence, and disabled-path safety.

Testing

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have tested this change with different models/inputs
• .venv/bin/python -m pytest tests/ -q — 5,206 passed, 55 skipped
• make format, make lint, and make format-check
• File-scoped pre-commit hooks, including Bandit and secret detection

Documentation

• I have updated the documentation accordingly
• I have added docstrings to new functions/classes
• I have updated the CHANGELOG.md

No user-facing documentation or changelog entry is needed for this internal opt-in hook; the public APIs and offset contracts are documented in code.

Code Quality

• I ran make format, make lint, and make format-check
• For Swift/OpenMedKit changes, I ran make format-swift and make lint-swift
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new warnings

Swift/OpenMedKit validation is not applicable because this PR changes only Python code and synthetic test data.

Dependencies

• I have not added any new dependencies
• OR I have added new dependencies and they are justified because: ____

Checklist

• I have read the contributing guidelines
• My commits have clear, descriptive messages
• I have squashed/organized my commits appropriately

Related Issues

Closes #1487

Screenshots/Examples

Not applicable; this change has no UI surface.