[tune] Reset Config for Trainables (#2831)

Adds the ability for trainables to reset their configurations during experiments. These changes in particular add the base functions to the trial_executor and trainable interfaces as well as giving the basic implementation on the PopulationBasedTraining scheduler.

Related issue number: #2741

python/ray/tune/trial_executor.py

@@ -106,6 +106,21 @@ class TrialExecutor(object):
         assert trial.status == Trial.PAUSED, trial.status
         self.start_trial(trial)
 
+    def reset_trial(self, trial, new_config, new_experiment_tag):
+        """Tries to invoke `Trainable.reset_config()` to reset trial.
+
+        Args:
+            trial (Trial): Trial to be reset.
+            new_config (dict): New configuration for Trial
+                trainable.
+            new_experiment_tag (str): New experiment name
+                for trial.
+
+        Returns:
+            True if `reset_config` is successful else False.
+        """
+        raise NotImplementedError
+
     def get_running_trials(self):
         """Returns all running trials."""
         raise NotImplementedError("Subclasses of TrialExecutor must provide "