Gizmo does not work with reversed z projection matrix

[YouSafe]

Using a reversed z projection matrix leads to the gizmo not being interactable. It might be a good idea to document this behaviour. In the meanwhile, this issue exists to potentially save someone time if they stumble upon the same issue.

[Azkellas]

I'm having no problem using egui-guizmo with an infinite reversed-z right handed projection matrix (computed via glam helpers). How are you using yours?

[YouSafe]

Tried it again, and now it's working for me too. Unfortunately, I lost the code that I used to try it out with but after trying different theories, I think it did not work for me because the matrix was wrong.

At the time of writing this issue, I also created a PR for my math library of choice: ultraviolet. That PR fixes the incorrect reverse-z-projection matrix (fu5ha/ultraviolet#168) and I might have just used some adapted version of the old and incorrect projection matrix for the gizmo.

This might be the code I used for generating the incorrect reverse-z projection matrix, which also reproduces the behaviour I got when I wrote this issue:

pub fn perspective_reversed_incorrect(
    vertical_fov: f32,
    aspect_ratio: f32,
    z_near: f32,
    z_far: f32,
) -> Mat4 {
    let t = (vertical_fov / 2.0).tan();
    let sy = 1.0 / t;
    let sx = sy / aspect_ratio;
    let nmf = z_near - z_far;
    Mat4::new(
        Vec4::new(sx, 0.0, 0.0, 0.0),
        Vec4::new(0.0, sy, 0.0, 0.0),
        Vec4::new(0.0, 0.0, z_far / nmf, -1.0),
        Vec4::new(0.0, 0.0, -z_near * z_far / nmf, 0.0),
    )
}

Changing the third column of the matrix in the following way makes the gizmo interactable again:

- Vec4::new(0.0, 0.0, z_far / nmf, -1.0),
+ Vec4::new(0.0, 0.0, -z_near / nmf, -1.0),

This matches the implementation of the current perspective_reversed_z_wgpu_dx_gl function (Note: $\frac{-z_n}{z_n-z_f}=\frac{-z_f}{z_n-z_f}-1$).

I apologize for wrongly assuming that this was an issue with the gizmo crate.