Massive refactor: split codegen into JSON and domain models

[Bromeon]

A lot of issues in godot-codegen have arisen because all business logic (analyzing symbols, generating code, ...) was operating on the raw JSON models from the extension API. While this worked well in the beginning, it caused more and more issues over time and became a scalability hazard:

• The JSON models had inconvenient data structures (e.g. Option<Vec<T>> instead of Vec<T>).
• Derived information needed to be constantly recomputed or stored elsewhere.
• Several bugs occurred due to mixing our Rust representation with Godot's original one, for example regarding names.
• The logic to map from JSON to domain model was all over the place -- centralizing it makes things much easier.

This is the foundation for more codegen cleanups and is direly needed for some of the features I plan to implement soon.

[GodotRust]

API docs are being generated and will be shortly available at: https://godot-rust.github.io/docs/gdext/pr-573