render_baselines not compatible with most recent myosuite update

[v9joshi]

With the most recent update to myosuite the observation and action steps work slightly differently (probably because of a different version of gym?). I had to make the following changes in render_baselines.py to get it to work.
line 20:
obs = env.obsdict2obsvec(env.obs_dict, env.obs_keys)[1]
This makes sure the input to the policy has the correct shape.

line 22:
obs, reward, done, *_ = env.step(action)

As env.step now returns 5 outputs instead of 4.