Fix Blog.MD wall-of-text by splitting megaparagraphs

Update Blog.MD

Submission: single notebook, restored Blog, eval comparison plot

Unify GRPO+eval in wildfire_train_eval_hf.ipynb; redirect legacy notebooks

Align HF notebooks with clone/pull setup; document eval in README and Blog.MD

Consolidate to two HF notebooks: wildfire_grpo_train_hf + wildfire_eval_plots_hf

Restore stashed work: notebooks, eval flags, plot helper, eval_policy wrapper

Sync submission notebooks + consolidated README, drop redundant READMEs

Document 20-iter deadline_v2_a10g as submission training run

Switch eval to OpenEnv WebSocket EnvClient, drop arbitrary step cap