fix: scale rewards to strict (0, 1) range and inject discoverable incident evidence tokens

fix: enforce true determinism and ground abstract action schemas

feat: complete final tier elite features (post-mortem, SLA, noise, and baseline ladder)

feat: implement elite scoring upgrades (reward shaping and adversarial Task 4)

feat: initial submission for OpenEnv hackathon - causalstream v3 verified