This PR significantly improves the dataset card for the SpatialGen Testset by:
- Adding `task_categories: ['image-to-3d', 'text-to-3d']` to the metadata for better discoverability.
- Including relevant `tags` such as `3d`, `scene-generation`, and `indoor-scenes`.
- Providing a concise introductory sentence based on the paper's abstract to give immediate context.
- Consolidating and expanding the usage instructions by integrating the inference code snippets (`infer_spatialgen_i2s.sh`, `infer_spatialgen_t2s.sh`) from the GitHub repository into a new "Sample Usage" section, alongside the existing visualization code.