Use SPiCiCAP dataset for future model training

@sandrinebedard pointed me to the following open-access dataset containing T2w (0.4x0.4x0.8mm) images for 22 healthy subjects:

https://data.mendeley.com/datasets/n2k7zz2xyt/1

@sandrinebedard applied the [dorsal rootlets model](https://github.com/ivadomed/model-spinal-rootlets/releases/tag/r20240730) on this dataset and rootlets segmetnation works pretty well!

Idea: we could use this dataset to enrich our training split for future model training!

Start by reviewing the SPiCiCAP dataset and the linked dorsal rootlets model release to understand the available T2w images and current segmentation results. Then identify how training data is organized in this repository and define how the 22 subjects would be added to a future training split. Done means the dataset integration and its effect on training are specified and reproducible.

Written by the indexing model from the issue text.