feat(integration): add DGL integration

[jeremy-wayland]

This pull request adds comprehensive support for DGL (Deep Graph Library) integration to the RINGS evaluation framework, enabling users to apply RINGS perturbations to DGL graphs in the same way as PyG (PyTorch Geometric) graphs. The changes introduce DGL-compatible wrappers for all core perturbations, update documentation and installation instructions for optional integrations, and add thorough tests to ensure correct behavior and compatibility.

DGL Integration:

• Added a new module rings/integrations/dgl.py that provides DGL wrappers (DGLOriginal, DGLEmptyFeatures, DGLRandomFeatures, DGLEmptyGraph, DGLCompleteGraph, DGLRandomGraph) for all core RINGS perturbations, enabling round-trip conversion between DGL and PyG graph formats.
• Updated rings/integrations/__init__.py to expose these DGL wrappers and ensure lazy loading of DGL dependencies.
• Modified SeparabilityStudy.apply to support DGL DGLGraph objects in addition to PyG Data/Dataset.

Documentation and Installation:

• Updated pyproject.toml to add DGL and Lightning as optional dependencies, and defined integration dependency groups for flexible installation. [1] [2]
• Expanded installation and usage instructions in README.md and Sphinx docs to cover optional integrations and DGL usage, including example code and dependency group usage. [1] [2] [3] [4] [5]
• Documented DGL wrappers and their usage in the integrations documentation, including API references.

Testing:

• Added tests/test_dgl_integration.py with comprehensive unit tests for all DGL perturbation wrappers, round-trip conversions, feature handling, and integration with SeparabilityStudy.
• Updated Lightning integration tests in tests/test_lightning_framework_integration.py to include DGL-based graphs and ensure compatibility with Lightning callbacks and studies.