Save model backbone+decoder name in the Model field and the dataset as well

[blumenstiel]

Currently, Model and Dataset are empty. Would be good to save them, either via a extra filed in the yaml or from the yaml values.

The dataset name is not straigtforward to extract from the yaml, e.g. when the generic datamodule is used. However, I think it is fair to assume that the task name is related to the dataset. Maybe it is best, if the user can specify the dataset name for each task and we use the task name as a fallback.