fixing bugs found with TorchModel().evaluate()

[kathryn-baker]

Fixing the following errors:

• native integer types from Python not recognised or utilised
• TypeError not raised with failing integer type causing evaluate() to fail silently instead of raising an error
• IndexError raised by evaluate() due to uneven dimensions of the default_tensor:

Traceback (most recent call last):
  File "User\lume-deployment\test.py", line 66, in <module>
    outputs = lumemodel.evaluate(input_dict)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "User\AppData\Local\anaconda3\envs\mlflow\Lib\site-packages\lume_model\models\torch_model.py", line 123, in evaluate
    input_tensor = self._arrange_inputs(formatted_inputs)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "User\AppData\Local\anaconda3\envs\mlflow\Lib\site-packages\lume_model\models\torch_model.py", line 279, in _arrange_inputs      
    input_tensor = torch.tile(default_tensor, dims=(*input_shapes[0], 1))
                                                     ~~~~~~~~~~~~^^^
IndexError: list index out of range

Also some odd formatting changes from my auto-formatter