QASC dataset: answerKey missing in test split

Short description
In QASC dataset, answerKey field is missing in the test samples.

Environment information

• Operating System: Ubuntu
• Python version: 3.7
• tensorflow-datasets/tfds-nightly version: 3.2.1
• tensorflow/tf-nightly version: 3.2.1.dev202009090105

Reproduction instructions

import tensorflow_datasets as tfds
import tensorflow as tf

tf.enable_eager_execution()
ds = tfds.load('qasc', split="test", shuffle_files=True)

for datapoint in ds:
    facts = [datapoint["fact1"].numpy().decode("utf-8"), datapoint["fact2"].numpy().decode("utf-8")]
    question = datapoint["question"].numpy().decode("utf-8")
    choices = {label.decode("utf-8"): text.decode("utf-8")
                for label, text in zip(datapoint["choices"]["label"].numpy(),
                                        datapoint["choices"]["text"].numpy())}
    answer = datapoint["answerKey"].numpy().decode("utf-8")
    print(answer)

Reproduce the issue by loading the QASC test split with the provided Python snippet and inspecting the dataset definition for the test samples. Check whether answerKey is included in the exposed test features, then rerun the snippet to confirm that every test sample provides it.