Require pillow and use KellyCoinflip instead of FrozenLake for gym te…

…st. These changes need to go together because the build is broke due to both reasons in the past 12-24 hours.

- Pillow was likely (I have not looked into that yet) being installed by another required package and is no longer a dependency of that package and thus the failure began.

- OpenAI Gym is not longer loading FrozenLake-v0 and FrozenLake-v1 is new and might not be available to all users. Moving the kwargs test to use KellyCoinflip resolves the problem. Here is what happened with FrozenLake-v0: openai/gym#2205 and openai/gym#2315

PiperOrigin-RevId: 391328529
Change-Id: I09e6e962a32330b24b1fc44ebe222fb1d842d5c3

setup.py

@@ -174,6 +174,7 @@ def get_required_packages():
       'gin-config >= 0.4.0',
       'gym >= 0.17.0',
       'numpy >= 1.13.3',
+      'pillow',
       'six >= 1.10.0',
       'protobuf >= 3.11.3',
       'wrapt >= 1.11.1',

tf_agents/environments/suite_gym_test.py

@@ -69,11 +69,12 @@ def testGinConfig(self):
     self.assertIsInstance(env, wrappers.TimeLimit)
 
   def test_gym_kwargs_argument(self):
-    env = suite_gym.load('FrozenLake-v0', gym_kwargs={'map_name': '4x4'})
-    self.assertTupleEqual(env.unwrapped.desc.shape, (4, 4))
+    env = suite_gym.load('KellyCoinflip-v0', gym_kwargs={'initial_wealth': 21})
+    self.assertEqual(env.unwrapped.initial_wealth, 21)
 
-    env = suite_gym.load('FrozenLake-v0', gym_kwargs={'map_name': '8x8'})
-    self.assertTupleEqual(env.unwrapped.desc.shape, (8, 8))
+    env = suite_gym.load('KellyCoinflip-v0',
+                         gym_kwargs={'initial_wealth': 50})
+    self.assertEqual(env.unwrapped.initial_wealth, 50)
 
   def test_load_gym_render_kwargs(self):
     env = suite_gym.load('CartPole-v1',