[cschanhniem]

Implements S1 from NEXT_100_PR_MAP.

What

• src/openamp_foundry/evidence/score_enemy_gap_report.py: SEG- schema; ScoreEnemyGapReport dataclass (16 fields) + CheapEnemyResult helper; validate_score_enemy_gap_report(); build_score_enemy_gap_report() (auto-computes verdict, best_enemy, gap, enforces dry_lab_only=True); format_score_enemy_gap_report()
• VALID_GAP_VERDICTS (4): gap_meaningful, gap_marginal, gap_absent, comparison_not_run
• VALID_CHEAP_ENEMY_TYPES (8): charge_score, length_score, hydrophobicity_score, isoelectric_point_score, aromaticity_score, combined_charge_length, random_baseline, sequence_similarity_to_known
• VALID_GAP_ASSESSMENT_METHODS (5): absolute_difference, relative_margin, rank_correlation, precision_at_k_delta, auc_delta
• MINIMUM_MEANINGFUL_GAP = 0.05 (absolute score units)
• tests/evidence/test_score_enemy_gap_report.py: 53 tests in 7 classes
• BASELINE: 9215 → 9278

Disconfirming pass

• dry_lab_only=False blocked (tested)
• TOY- candidate_family_id blocked (tested)
• gap_meaningful requires gap >= MINIMUM_MEANINGFUL_GAP (tested)
• gap_absent requires gap <= 0 (tested)
• comparison_not_run requires n_enemies_tested == 0 (tested)
• gap inconsistent with scores blocked (tested)
• n_enemies_tested must match len(enemy_results) (tested)
• invalid enemy_type blocked in results (tested)
• build() picks highest-scoring enemy as best (tested)