Tokenization error for ReCoRD

[claravania]

Describe the bug
Error when tokenizing with roberta-large.

ReCoRDTask
  [train]: /scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/train.jsonl
  [test]: /scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/test.jsonl
  [val]: /scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/val.jsonl
Traceback (most recent call last):
  File "/scratch/cv50/jiant-2/jiant/jiant/proj/main/tokenize_and_cache.py", line 214, in <module>
    main(args=RunConfiguration.run_cli_json_prepend())
  File "/scratch/cv50/jiant-2/jiant/jiant/proj/main/tokenize_and_cache.py", line 165, in main
    examples=task.get_train_examples(),
  File "/scratch/cv50/jiant-2/jiant/jiant/tasks/lib/record.py", line 122, in get_train_examples
    return self._create_examples(lines=read_json_lines(self.train_path), set_type="train")
  File "/scratch/cv50/jiant-2/jiant/jiant/tasks/lib/record.py", line 134, in _create_examples
    passage_text = line["passage"]["text"]
TypeError: string indices must be integers

To Reproduce

1. commit: aa4d111
2. Describe the environment where you're using jiant, e.g, "2 P40 GPUs": 1 P40 GPU
3. Provide the experiment config artifact (e.g., defaults.conf):

{
  "task": "record",
  "paths": {
    "train": "/scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/train.jsonl",
    "test": "/scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/test.jsonl",
    "val": "/scratch/bowman/IRT_Experiments/jiant-2/experiments/tasks/data/record/val.jsonl"
  },
  "name": "record"
}