Support Custom User Metrics

[EdwinB12]

• Allow the user to write their own metrics (adding general torchmetrics support #883)
• Enable support for custom metrics to be added to the list of tuning and evaluation metrics (Update API for getting metric config #912, Enable custom metrics in AutoEmulate compare loop #913)
• Add support for serializing and deserializing custom metrics
• Add API to support custom scoring across tuning folds

[sgreenbury]

There are a couple of places that could be good to support additional metrics:

• autoemulate/autoemulate/core/compare.py

  Line 507 in a04fa07

  best_result = self.best_result()

• autoemulate/autoemulate/core/compare.py

  Line 391 in a04fa07

  np.mean(score).item() for score in scores

  - this one relates to Update metric choosing optimal hyperparameters #661 where mean plus one standard error was explored