Annotation file in different folder than the txt file for BRAT parser

As far as I understand, the current BRAT parser implementation assumes that the annotation file is stored in the same folder as the `.txt` file.

https://github.com/bigscience-workshop/biomedical/blob/db763955a64cabe8fe30700c30d2dbb017fc995f/utils/parsing.py#L149

This is not true for the [EBM-NLP](https://github.com/bepnye/EBM-NLP) dataset.

As I only found one such dataset, this is probably a very low priority, but I wanted to ask if it would be worth refactoring the parser at some point so the `annotation_file` could be parametrised?