feat(M1): confidence scoring stubs — claims-native, update_claim_confidence raises NotImplementedError

[ARC345]

Summary

Replace the paper-centric confidence module with claims-native stubs. Full scoring model is deferred to M2.

What changes

• get_initial_confidence_for_claim(claim_type) — returns per-type initial confidence:
  • empirical: 0.3 (requires evidence)
  • methodological: 0.4
  • conceptual: 0.5
  • boundary: 0.4
• update_claim_confidence(claim_id, stance, weight, source_id) — raises NotImplementedError with a detailed message pointing to the beta implementation plan

Files

• backend/scoring/confidence.py — rewrite
• backend/tests/test_scoring.py — rewrite