[cschanhniem]

Implements R5 from NEXT_100_PR_MAP — closes Phase R (scientific output readiness).

What

• src/openamp_foundry/evidence/preprint_evidence_bundle.py: PEB- schema; PreprintEvidenceBundle dataclass (19 fields); validate_preprint_evidence_bundle(); build_preprint_evidence_bundle() (enforces canonical DRY_LAB_ONLY_DISCLAIMER, dry_lab_only=True); format_preprint_evidence_bundle()
• VALID_BUNDLE_STATUSES (5): draft, internal_review, approved, submitted, published
• VALID_PREPRINT_SERVERS (7): bioRxiv, chemRxiv, Research Square, SSRN, Zenodo, OSF Preprints, not_submitted
• VALID_CLAIM_STRENGTH_TIERS (4): computational_nomination, validated_computational, preliminary_wet_lab, replicated_wet_lab
• DRY_LAB_ONLY_DISCLAIMER: canonical anti-overclaim statement enforced in every bundle
• tests/evidence/test_preprint_evidence_bundle.py: 66 tests in 5 classes
• BASELINE: 9152 → 9215

Disconfirming pass

• dry_lab_only=False blocked (tested)
• TOY- family ID blocked (tested)
• wrong DRY_LAB_ONLY_DISCLAIMER blocked (tested) — canonical text is mandatory
• published requires preprint_doi (tested)
• draft blocks preprint_doi (tested)
• submitted/published require real server (not 'not_submitted') (tested)
• replicated_wet_lab requires n_confirmed_hits >= 2 (tested)
• empty artifact_ids blocked (tested)