Example of migrating backends to ApplicationHandler

[madsmtm]

See #3386.

I stubled into a roadblock for this, as there are several backends which are currently queuing events. This means they either need some sort of event structure themselves if we're to remove Event<T>, or will have to be rewritten to not queue events (preferred).

Backends that are currently queuing events: macOS, iOS, web, Windows, Wayland.

I can only speak for macOS and iOS, but I'm pretty sure those already prevent re-entrancy for most events, so migrating to handling events immediately will be the strictly better solution.

In any case, this PR showcases the migration of the Redox backend, so it should be technologically possible.

[kchibisov]

yeah, I can't write like that on Wayland at all until I change other API. The issue is that Window creation must be made async before I can do so, because I do dispatch events and I can get events during creation of the window for other windows.

[madsmtm]

Right, I missed the EventSink when I reviewed which backends are queuing events.

[madsmtm]

If it's not really feasible to rewrite the Wayland backend to utilize this trait until we do other work, then... Don't! Use the crate::event_helper::map_event approach like I've done for macOS currently.

I'm also only talking about removing the Event<T> from public API; it could stay in private APIs for backends that need it to queue certain events, until we slowly get rid of unnecessary queuing.

[kchibisov]

@madsmtm I may dispatch events from the window creation, like I'd need to change entire window creation process to be async like I have on my branch.