Apply suggestions from code review

Author: Borda

src/lightning/pytorch/plugins/precision/fsdp.py

@@ -86,6 +86,7 @@ def clip_grad_by_norm(self, module: Optional[Module], optimizer: Optimizer, clip
         # see https://pytorch.org/docs/stable/fsdp.html#torch.distributed.fsdp.FullyShardedDataParallel.clip_grad_norm_
         if module is None:
             return
+        assert isinstance(module.clip_grad_norm_, Module)
         module.clip_grad_norm_(clip_val)
 
     @property