[rllib] Support 1-dimensional action spaces (PPO) (ray-project#1347)

* Small fix for supporting custom preprocessors

* PEP8

* Remove squeeze from actions

python/ray/rllib/ppo/runner.py

@@ -155,7 +155,7 @@ def load_data(self, trajectories, full_trace):
             [trajectories["observations"],
              trajectories["value_targets"] if use_gae else dummy,
              trajectories["advantages"],
-             trajectories["actions"].squeeze(),
+             trajectories["actions"],
              trajectories["logprobs"],
              trajectories["vf_preds"] if use_gae else dummy],
             full_trace=full_trace)