refactor : grad_norm 指标简化，clip_grad_norm 移至 executor

- metrics 默认加入 grad_norm，移除 grad_std/max/min/mean/nan_num
- grad_norm 默认返回总 L2 范数，per_param=True 返回各参数范数
- clip_grad_norm 从 callback 移至 BaseExecutor/FSDPExecutor
- FSDPExecutor 覆盖为 model.clip_grad_norm_() 保证分布式正确
- ctx_get_grad_norm 改为读取 context.grad_norm

astrai/config/train_config.py

@@ -72,7 +72,7 @@ class TrainConfig(BaseConfig):
         metadata={"help": "Number of batch iterations between metric logs."},
     )
     metrics: List[str] = field(
-        default_factory=lambda: ["loss", "lr"],
+        default_factory=lambda: ["loss", "lr", "grad_norm"],
         metadata={"help": "Metrics to record during training."},
     )