Restore moderation_benchmark.json from d741d4b (128 scenarios, 100/100 checks)

2b 3a

Appeal mechanic: is_adversarial + env.appeal() 2-turn flow (92/92 checks)

Fill easy GT gaps: full label×action coverage (79/79 checks)

Cross-post campaign mechanic: campaign_id in state, +0.15 escalate-all bonus (61/61 checks)

10 ambiguous hard scenarios + full valid_actions test suite (53/53 checks)

Close all 4 scoring gaps (+~6 pts)

phase 1 and phase 2a

ContentModerationEnv v1.0 — complete OpenEnv benchmark