Expose matched index pairs as NearestNeighborMatch.matched_indexes_ (#621)

[jeongyoonlee]

Proposed changes

Expose the matched index pairs computed inside NearestNeighborMatch.match() as a fitted attribute matched_indexes_ — a two-column pandas.DataFrame (from, to) where each row is one matched pair. Useful for joining matched pairs back to upstream metadata or auditing the matching outcome without re-running the algorithm.

Previously these (from, to) pairs were computed and then discarded; only the joined DataFrame was returned. Under ratio > 1 a single from index can appear multiple times against distinct to indices, so the attribute captures the pair mapping (not just the deduplicated from-set). Both the replacement (NearestNeighbors) and no-replacement (caliper-loop) paths are covered. match_by_group calls match() per group, so the attribute reflects the most recent call.

Fixes #621.

Credit

Reimplements the work of @jbbqqf in #897, which is CLA-blocked and cannot be merged directly. Full credit to @jbbqqf for the original implementation and analysis; this PR re-lands the same feature against current master under a CLA-clean author.

Types of changes

• New feature (non-breaking change which adds functionality)

Tests

• New test_nearest_neighbor_match_exposes_matched_indexes in tests/test_match.py covering both branches (schema, consistency with the returned matched frame, and pair-level granularity under replace=True, ratio=2). Fails on master with AttributeError, passes here.
• Full tests/test_match.py — 6 passed.