Text this: A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others