Fixes panic when using Option<Gizmos>

[kfc35]

Objective

• Fixes Optional SystemParam (especially gizmos) #14949
• Currently an Option<Gizmos> causes a panic when the config group backing Gizmos has not been initialized. The desired behavior is to have it be None instead of course.

Solution

Gizmos is a non #[derive]'d SystemParam that requires its

• GizmoConfigGroup to have an entry in GizmoConfigStore
• Various GizmoStorages to be inserted as resources.

In order to support a non-panicking Option<Gizmos> SystemParam in a non-intrusive way for users when the Gizmo Config Group has not been set up intentionally, this PR:

• Adds ergonomic methods to get a group’s config fallibly to the GizmoConfigStore
• Changes the validate_param method of Gizmos to return a skippable error in the event that it could not find its own config in the GizmoConfigStore
• Changes the apply method of GizmoBuffer to only apply changes if the GizmoStorage for the config group exists.

Testing

I modified the 2d gizmos example (cargo run --example 2d_gizmos) by:

• commenting out the init of the custom gizmo group // .init_gizmo_group::<MyRoundGizmos>()
• Using options (i.e. Option<Gizmos> and Option<Gizmos<MyRoundGizmos>>) for the two Gizmos system parameters in draw_example_collection and requiring my_gizmos_option.is_some() before using an unwrapped my_gizmos to draw things.
• Using the new GizmoConfigStore method try_config_mut when trying to get the config for MyRoundGizmos, and only proceeding to change its settings if my_config_option.is_some()

There are no crashes. Default Gizmos continue to be drawn, and MyRoundGizmos are not drawn. Changing the settings still works for default gizmos, and has no effect if they were to be applied to MyRoundGizmos. There seems to be no deterioration in performance for the example.

[kfc35]

Note to self that I need to also make this change to queue() #22800