calculate_results appears to only use the first correct answer, not all correct answers

[kpounder]

Describe the bug
In reviewing the code for calculate_result for QuestionAnsweringModel, I saw this line

simpletransformers/simpletransformers/question_answering/question_answering_model.py

Line 1416 in 76d9801

truth_dict[answer["id"]] = answer["answers"][0]["text"]

which made me think that for evaluation we're only considering the first correct answer. However, according to the docs here

https://simpletransformers.ai/docs/qa-data-formats/#evaluation-data-format

it looks like there can be multiple correct answers for a given question.

{
                "id": "00001",
                "is_impossible": False,
                "question": "Where does the series take place?",
                "answers": [
                    {
                        "text": "region called the Final Empire",
                        "answer_start": 38,
                    },
                    {
                        "text": "world called Scadrial",
                        "answer_start": 74,
                    },
                ],
            }

Am I interpreting the code correctly? Is this a bug?

To Reproduce
N/A

Expected behavior
Take "maximum" across right answers -- i.e., compare given answer to each of right answers, and if any one is correct, then correct; else, if any one is similar, then similar; else incorrect.

Screenshots
N/A

Desktop (please complete the following information):
N/A

Additional context
N/A