fix: token-level ratio and prompt masking in GRPO strategy

- Mask prompt tokens to 0 so their logprobs excluded from ratio/KL
- Switch to token-level ratio + PPO clipping via reduction='none'
- Slice response token logprobs from full sequence output
- Replace k3 KL estimator with non-negative k1 estimator
- Fix epsilon from finifo.eps (~1e-38) to 1e-8
- Remove unused 'reduction' param from GRPOStrategy.__init__
- Clarify offline batch semantics in docstring
- Add 11 unit tests for masking, advantage, KL, sync, clipping
- Sync training.md and architecture.md docs

assets/docs/architecture.md

@@ -499,7 +499,6 @@ classDiagram
             +float clip_eps
             +float kl_coef
             +int group_size
-            +str reduction
             +int sync_interval
             +compute_loss(batch) Tensor
             +sync_ref_model()