The JSON datasets must include a reference to the resultTree

We can add a symlink to improve the traceability of every dataset, as it connects directly to the experiment. I also prevents the GC from removing any experiment dependency as they will be runtime dependencies of the article.