✨ Upgrade SQLAlchemy to 2.0, including initial work by @farahats9

This includes @farahats9 commits from #563 and builds on top.

• I reverted a few things, e.g. altering the docs to workaround for a previous lack of support for forward references.
• Added support/fix for forward reference errors.
• Removed several sections/modules that are no longer necessary, as were mainly to add type annotations to SQLAlchemy things, but now with SQLAlchemy's new proper types that's no longer necessary and I can reuse SQLAlchemy's things directly.
• Updated all needed tests, e.g. to support the latest FastAPI.
• Updated all the generated selects to use the new SQLAlchemy types.
• Updated the classes and type annotations with the new SQLAlchemy types, with some overloads to make sure they are compatible with SQLModel. E.g. select().where().
• Redefine all functions that would take column operations to accept the types used by SQLModel, e.g. or_(Hero.name == "Deadpond") as SQLAlchemy's version would require Hero.name to be a Mapped[str] instead of a str.
• I explored a lot if it was possible/sensible to use and encourage the new Mapped[X], but sadly that breaks type annotations for creating new instances and dataclass_transform, as then that would expect to receive values with Mapped[X], e.g. Mapped[str] instead of str. And it would also add complications to the Pydantic handling. So, this was a lot of work mostly exploratory, but not much ended in the PR.