Fixed up the documentation for tf.contrib.layers.batch_norm and...

Fixed up the documentation for tf.contrib.layers.batch_norm and tf.layers.batch_normalization explaining how to include the update ops in the train_op. Also simplified the existing tf.layers.batch_normalization documentation to the least error-prone option (adding to the train_op). (#10308)

PiperOrigin-RevId: 156609483