Fix exercise 1.3

[lbertge]

Actions are interpreted as scalars. In NumPy this is represented as an array, but with one of its dimensions size 0. Here, we desire a vector with the shape (act_dim, ). In the get_action() method, the action is assigned to the result of sess.run(pi ...) which returns an vector with shape (act_dim, 1). We add the [0] here to remove the second dimension so it can be used properly. Without it, we can get an error, like the following:

> python exercise1_3.py --env RoboschoolHalfCheetah-v1
...
  File "/home/anaconda3/envs/spinningup/lib/python3.6/site-packages/roboschool/gym_forward_walker.py", line 43, in apply_action
    j.set_motor_torque( self.power*j.power_coef*float(np.clip(a[n], -1, +1)) )
TypeError: only size-1 arrays can be converted to Python scalars

[lbertge]

Sorry, just noticed this is a duplicate of #106.