[RLlib] Fix use_lstm flag for ModelV2 (w/o ModelV1 wrapping) and add it for PyTorch. (#8734)

doc/source/rllib-models.rst

@@ -121,7 +121,7 @@ Once implemented, the model can then be registered and used in place of a built-
     from ray.rllib.models import ModelCatalog
     from ray.rllib.models.torch.torch_modelv2 import TorchModelV2
 
-    class CustomTorchModel(nn.Module, TorchModelV2):
+    class CustomTorchModel(TorchModelV2):
         def __init__(self, obs_space, action_space, num_outputs, model_config, name): ...
         def forward(self, input_dict, state, seq_lens): ...
         def value_function(self): ...