indexer: emit typed call/import edges with enclosing-symbol attribution (Python)

[IceRhymers]

Part of #82. Depends on #83.

Summary

Extend the tree-sitter indexing walk to emit typed call and import edges with enclosing-symbol attribution for Python, and write them through the existing per-file lifecycle.

Scope

• Extend indexer/languages.py with a per-language edge node map alongside SYMBOL_KINDS (same single-source-of-truth architecture); Python first: call (rightmost identifier of the callee) and import_statement/import_from_statement.
• Extend the walk in indexer/symbols.py (or a sibling edges extractor sharing the walk) to emit edges in the same pass, tracking the enclosing definition node on the traversal path — no second walk, no join at write time.
• Write edges in indexer/store.py with the same delete-and-reinsert-per-file pattern as symbols, inside the existing transaction (pure DML); ensure branch sweep/orphan cleanup covers edge rows via FK cascade.
• Bump INDEX_SEMANTICS_VERSION so the CAS skip logic forces re-extraction.
• Unit fixtures mirroring tests/unit/test_symbols.py: nested calls, method vs bare calls, aliased imports, edge line numbers (1-based).

Acceptance criteria

• Python files produce typed call/import edges with correct enclosing-symbol attribution and line numbers.
• Edge extraction overhead ≤ ~15% of parse+walk (deep-dive probe measured 12%).
• Delete-and-reinsert + sweep proven by tests (re-index removes stale edges; branch retirement cascades).
• Single deliberate INDEX_SEMANTICS_VERSION bump.

[IceRhymers]

Completed via child PR #91, squash-merged into the #82 integration branch (integration/knowledge-graph-reference-edges). The umbrella draft PR #89 remains open until the full epic is integrated and human-reviewed into protected master.