deploy: fresh upload via HF API — force rebuild with latest code

feat: curriculum training, 4 reward funcs, session API, preflight tests, compare_inference

added inference comparisor

feat(rl): Add Unsloth + TRL GRPO training pipeline scaffolding