Properly log model artifact in MLFlow

Nobody has claimed this yet.

The model weights file is logged automatically in the primary results directory, MLFlow offers a similar method to store models, but expects those logged models to include either a `signature` or `input_example`. Neither of which seems to be working as expected within the context of Fibad. Providing nothing results in a `WARN` log, and attempting to provide example input produces other error messages.

```
with mlflow.start_run(log_system_metrics=True):
_log_params(config)

# Run the training process
trainer.run(train_data_loader, max_epochs=config["train"]["epochs"])

mlflow.pytorch.log_model(model, "models") # This should include either signature = ... or input_example = ...
```

Start from the MLflow run block shown in the issue, especially the trainer.run call and mlflow.pytorch.log_model invocation. Reproduce the WARN and input_example errors with the existing training configuration, then trace how the trained model and example inputs are represented in Fibad. Done means model artifacts are logged without the warning or reported errors and include the required metadata.

Written by the indexing model from the issue text.