Provide the projection and view matrices separately in the 3D camera, or add a screen-to-world method

[Alexandre425]

Mouse picking, the process of picking an entity or point in 3D space with your mouse, requires that the projection and view matrices be separate.

Currently, these two are available only together, through the matrix() method:

macroquad/src/camera.rs

Lines 209 to 225 in bd00be0

	fn matrix(&self) -> Mat4 {
	let aspect = self.aspect.unwrap_or(screen_width() / screen_height());
	match self.projection {
	Projection::Perspective => {
	Mat4::perspective_rh_gl(self.fovy, aspect, Self::Z_NEAR, Self::Z_FAR)
	* Mat4::look_at_rh(self.position, self.target, self.up)
	}
	Projection::Orthographics => {
	let top = self.fovy / 2.0;
	let right = top * aspect;
	Mat4::orthographic_rh_gl(-right, right, -top, top, Self::Z_NEAR, Self::Z_FAR)
	* Mat4::look_at_rh(self.position, self.target, self.up)
	}
	}
	}

If they were provided separately, it would be much easier to get the ray corresponding to a specific point on the screen. Alternatively, the entire operation (pictured below) could be provided as a method, returning the ray in world space:

// Ray in eye/view space
let ray_eye = projection_matrix.inverse() * vec4(screen_pos.x, screen_pos.y, -1., 1.);
// Ray in world space
let ray_world = view_matrix.inverse() * vec4(ray_eye.x, ray_eye.y, -1., 0.).normalize();