feat(cat5): component-size distribution and non-trivial summaries

[jphein]

Summary

Closes #17.

• Adds component_size_distribution: dict[str, int] to GapDetectionReport — histogram buckets "1", "2-5", "6-20", ">20" computed during the existing component iteration
• Adds non_trivial_components: list[dict] — each entry has "size" and "types" (sorted entity_type list) for components with ≥2 nodes
• format_report() renders both: distribution as a compact line, non-trivial components as a bulleted list

Test plan

• test_component_size_distribution — verifies bucket counts match fixture ground truth
• test_non_trivial_components — verifies list length and structure
• test_format_report_shows_distribution — verifies rendered output contains bucket labels
• Existing Cat 5 tests pass unchanged

🫏 Generated with Claude Code

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[M0nkeyFl0wer]

Ship. The bucket scheme (1, 2-5, 6-20, >20) is well-chosen — log-spaced enough to surface a long tail without exploding into 50 unique counts on a sparse graph — and the per-non-trivial-component type composition is the right next layer of detail. Computed during the existing component iteration so it adds zero structural cost. Clean.