[brief] Make the train_step function not abstract.

[detailed]
- This actually makes the model base class more flexible because it
  means that now the user can create models that isolate each individual
  step (test/train/validate) as opposed to one that has to cover
  training even when it's not needed.

src/helios/model/model.py

@@ -211,7 +211,6 @@ def on_training_batch_start(self, state: TrainingState) -> None:
         """
         self._loss_items.clear()
 
-    @abc.abstractmethod
     def train_step(self, batch: typing.Any, state: TrainingState) -> None:
         """
         Perform a single training step.