Update/core docs

[jhnwu3]

This pull request significantly improves and expands the PyHealth documentation, providing clearer explanations, practical usage guidance, and new sections for key components. The updates aim to make it much easier for users—especially newcomers—to understand how to load data, define tasks, work with models, and use knowledge graphs in the PyHealth ecosystem.

Major documentation improvements:

General usage and data loading:

• Expanded the "datasets" documentation to explain how PyHealth loads data, the distinction between BaseDataset and SampleDataset, initialization parameters, how to use custom YAML schemas, and common pitfalls for new users.

Tasks and customization:

• Added a comprehensive section on writing custom tasks, including subclassing BaseTask, specifying input/output schemas, processor string keys, and how set_task() works with multiprocessing.

Models and architecture:

• Rewrote the "models" documentation to explain the relationship between processors, datasets, and models; provided a model selection table; described the BaseModel and EmbeddingModel classes; detailed task modes and loss functions; and included a template for building custom models.

Knowledge graph support:

• Added a new page documenting the pyhealth.graph module, covering what knowledge graphs are, how to construct and explore them, how to extract patient subgraphs, integration with tasks, and support for pre-computed node embeddings.

[jhnwu3]

change pauls email to the correct one in the about.rst and .toml