Support f32 directly in process and physics_process

[Bromeon]

It is now possible to write either of the following for process and physics_process:

#[godot_api]
impl INode for MyClass {
    // Continues to work: f64.
    fn process(&mut self, delta: f64) { ... }

    // Transparently cast to single-precision float:
    fn process(&mut self, delta: f32) {
        // Work with f32 here.
    }
}

New approach based on proc-macro transformation. It's a bit magic, but for now it's the only practical solution that doesn't put f64 users at a disadvantage. One viable alternative would have been adding extra process_64 + physics_process_64 overloads, but this would come at the cost of hundreds of extra generated methods. Another option would be to always use f32 and let f64 users query the delta via dedicated API.

Resolves #510. See this thread for discussion of the various approaches.

We need to see how this turns out, maybe this can change again in the future. Will definitely need doc updates.

---

This PR also took the opportunity to refactor some small parts in #[godot_api]. To skip those changes, just exclude the first commit from the diff.

[GodotRust]

API docs are being generated and will be shortly available at: https://godot-rust.github.io/docs/gdext/pr-1110