Timers for tuning and postprocessor for early stopping

Background

Other AutoML libraries use timers to limit the amount of time models take in training. I think this would be good to add here so tuning times don't take longer than 3 hours or something that the user can set.

Tasks

• Add timers for each model and/or in total where once the time is met, iterations are stopped for that model or in total, and the top models found thus far are returned. This should be something the user can override with a different time or possibly setting it to -1 to turn off timers.
• Changes are documented in documentation.

Definition of Done

The above functionality is implemented for classical and machine learning models.