feat(perf): RAM-first indexing — batch SQLite write at end of scan (closes #10 phase-1)

[Wolfvin]

Summary

Implements issue #10: RAM-first indexing with batch SQLite write.

What changed

• Collect graph_nodes / graph_edges in Python lists during scan (already in-memory from parsers; this PR makes the batch-write boundary explicit)
• Single batch write at end of scan: one executemany() for nodes + one for edges, wrapped in a BEGIN EXCLUSIVE transaction (isolation_level=None + explicit BEGIN EXCLUSIVE / COMMIT)
• Refactored populate_graph_tables (full scan) and incremental_graph_update (incremental scan) to collect all rows in memory BEFORE opening the EXCLUSIVE lock, minimizing lock-hold duration
• New --scan-stats flag prints a two-line timing breakdown to stderr:

  Scan stats: 1240 files, 3091 nodes, 29285 edges
  Index time: 2.3s (parse: 1.8s, write: 0.5s)
  

• --incremental still works — partial update path is preserved; the EXCLUSIVE-lock refactor applies to both full and incremental writes
• Output to stderr so default scan output (stdout JSON / formatted text) is byte-identical to the pre-[PERF] RAM-first indexing pipeline: index in-memory, flush once to SQLite #10 behavior (backward compat)

Verification

• Tests: 1008 passed, 87 skipped (12 new tests added in tests/test_scan_stats.py covering the --scan-stats flag, the BEGIN EXCLUSIVE batch-write atomicity, and the incremental-scan interaction)
• Backward compat: default codelens scan output unchanged (verified by subprocess test that asserts no Scan stats: / Index time: lines on stderr without --scan-stats)
• --incremental still works: existing tests/test_graph_incremental.py (18 tests) all pass
• File headers preserved on all touched files (scan.py, graph_model.py) and added on the new test file
• Pre-existing unrelated failure: tests/test_hybrid_engine.py::TestHybridIntegration::test_query_confidence_without_deep — verified to fail on main (commit 470e602) before this PR

Closes #10

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud