Improve lifetime insertions for #[pyproto]

[kngwyu]

This PR allows:

#[pyproto]
impl PyIterProtocol for Iter {
    fn __iter__(slf: PyRef<Self>) -> PyRef<Self> {
        slf
    }
}

[davidhewitt]

Thanks! This actually helps fix a lot of #1031!

I think this change needs to go a bit further before merging, because at the moment it fixes some protocol methods but not all of them. For example, this code is still not supported:

#[pyproto]
impl PyDescrProtocol for MyClass {
    fn __get__(
        mut slf: PyRefMut<Self>,     // <--- This is very similar to your example, but is TernaryS receiver
        _instance: PyRef<Self>,      // <--- This one still has a problem
        _owner: Option<&PyType>,     // <--- This reference is nested in an Option, so get_arg_ty misses
    ) -> PyRefMut<Self> {            // OK after this PR =)
        slf
    }

I was actually thinking about a refactoring the other day which refactors the MethodProto enum to a struct instead. I can't find my branch right now, but this was the essence of the change:

use crate::method::SelfType;

pub struct MethodProto {
    name: &'static str,
    proto: &'static str,
    receiver: SelfType,
    args: &[&'static str],         // names of all the arguments
    result: Option<&'static str>   // name of the return type, if any (I think only `Free` does not have)
}

Maybe taking that refactoring now would help this PR? I can submit as a different PR, or you can use this idea directly in this one?

[davidhewitt]

I think you should add this inside get_arg_ty instead of calling here.

[davidhewitt]

It would be nice if we could support adding lifetimes to references too. I notice that get_arg_ty does this.

[kngwyu]

Thanks, I made it more general. Now your descr example works too.

I was actually thinking about a refactoring the other day which refactors the MethodProto enum to a struct instead. I can't find my branch right now, but this was the essence of the change:

It sounds nice since the current implementation is crafty. However, I want to leave it for another chance, since I have to complete another PR first (for #844).

[davidhewitt]

Nicely done! I'm happy for that refactoring to wait for another time 👍