[Suggestion] some potentially other useful subsets for BeyondArena

These are some potentially useful subsets for the filter, as these data sets are common:

- Data sets with joint groups between train and test. There's already a subset with disjoint groups, so it would be easy to implement the opposite.
- Noisy data
- Sparse data

Start by tracing how the existing disjoint-groups subset is defined and exposed in the dataset-filtering code. Confirm the intended definitions for joint train/test groups, noisy data, and sparse data with the issue discussion; done means the requested subsets are available through the filter and covered by the project's existing checks.

Written by the indexing model from the issue text.