Add journey/comparison/demo charts + push real model card to HF Hub

Real GRPO results: Llama 3.2 3B avg 0.075 -> 0.482 (6.4x) on VAPT-Env

first commit