Show score breakdown and selection rationale in dry-run / preview output

[chigichan24]

Why

• The current dry-run output (scripts/cli.ts:177-186) only shows [score] label + keywords + session count.
• That tells the user what ranked highly, but not why. Without re-reading the algorithm doc, users cannot quickly judge whether the score is meaningful or noisy.
• A richer breakdown would let users immediately see which signal drove the ranking and whether to trust it. This is part of making the dry-run / preview feel "rich enough to decide on" rather than "barely a teaser".

What

• For each top candidate in dry-run / preview output, surface:
  • Per-signal contribution to reusabilityScore (text TF-IDF, tool-IDF, structural features, successRate, helpfulness, etc.)
  • Top connected sessions with brief summaries (use SessionSummary.summaryText when available)
  • Top connected topics from the knowledge graph (community membership, centrality)
• Output should stay terse and human-readable; design for skim-ability, not exhaustive logs.

How

• ReusabilityScore already carries some component fields (successRate?, helpfulness?). Extend it to expose explicit per-signal contributions used during ranking.
• Centralize the per-candidate detail rendering in a small helper so dry-run and preview share the same source of truth.
• Tests: snapshot-style assertions on the rendered string for a couple of fixture candidates.

Notes

• Independent from Add an automatic Skill evaluation pipeline (skill-creator-style quality checks) #20 (Skill evaluator) but complementary: once Add an automatic Skill evaluation pipeline (skill-creator-style quality checks) #20 lands, the evaluator score becomes another row in this breakdown.