feat(training): outcome-based RLVR reward + clean Colab T4 notebook

feat(training): SFT dataset + stall detection in eval rollout

fix(training): per-action reward scoring vs heuristic oracle

feat: enhance completion parsing to handle truncated JSON and `<think>` blocks

feat: add per-family evaluation and plotting for training curves

feat: Add training curve evaluation and plotting utilities with unit tests

Add training notebook and benchmark runner for ChargebackOps