Add DocVQA

## Adding a Dataset
- **Name:** DocVQA
- **Description:** Document Visual Question Answering (DocVQA) seeks to inspire a “purpose-driven” point of view in Document Analysis and Recognition research, where the document content is extracted and used to respond to high-level tasks defined by the human consumers of this information.
- **Paper:** https://arxiv.org/abs/2007.00398
- **Data:** https://www.docvqa.org/datasets/docvqa
- **Motivation:** Models like LayoutLM and Donut in the Transformers library are fine-tuned on DocVQA. Would be very handy to directly load this dataset from the hub.

Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/main/ADD_NEW_DATASET.md).