Add DGES algorithm and enhance GES with updated BIC score and colored graph visualization

[lokali]

• Add DGES (Deterministic GES): Algorithm 1 from Li et al. (NeurIPS 2024) for causal discovery in the presence of deterministic relations. Three-phase approach: MinDC detection, DC-aware GES, optional exact search.
• Add deterministic-aware BIC score (local_score_BIC_from_cov_deterministic) that handles zero residual variance gracefully.
• Add covariance-based BIC score (local_score_BIC_from_cov) for GES, enabling GES to run from precomputed covariance matrices.
• Add lambda_value parameter to GES for controlling BIC penalty strength.
• Add node_names parameter to GES for custom node labeling.
• Add GraphUtils.plot_graph() for rendering colored causal graphs with manual per-node colors or automatic category-based coloring.
• Add GraphUtils.get_category_colors() helper for category-to-color mapping.
• Update documentation: DGES page, enhanced GES page, Datasets page, Graph Visualization page under Utilities.

[MarkDana]

Thank you so much @lokali for the efforts. Wonderful job in 1) resolving previous issues on lambda penalty term, 2) speedup >5x, and many more!

I have met with Loka offline and checked the code part; it looks good to me, except for a quick question above on the kernel based score.

@kunwuz could you please then review the readthedocs html part, and see if they render properly? If so I think this pr is ready to be merged.

Loka thank you again!

[kunwuz]

Thanks for the PR and review! The doc part is a bit different--we previously only edited/added rst files like docs/source/search_methods_index/Score-based causal discovery methods/GES.rst, and building html is done online via readthedocs. Could you please remove the html files here, and open a new pr with new rst files similar to the one pointed above?

[lokali]

@MarkDana @kunwuz Thanks for your reply! I have update the code in rst, and removed those in html. Please kindly review it!