Change readme example to explicitly set shuffle_files

PiperOrigin-RevId: 279801514

Author: Ryan Sepassi

README.md

@@ -53,7 +53,7 @@ tf.compat.v1.enable_eager_execution()
 print(tfds.list_builders())
 
 # Construct a tf.data.Dataset
-ds_train, ds_test = tfds.load(name="mnist", split=["train", "test"])
+ds_train = tfds.load(name="mnist", split="train", shuffle_files=True)
 
 # Build your input pipeline
 ds_train = ds_train.shuffle(1000).batch(128).prefetch(10)