As discussed in the Plasma sprint.
Removes some boilerplate code and turns it into C++ overrides.

As some background:

At a recent Plasma sprint, one comment was that adding protocols to KWayland was an absolute pain involving writing a tonne of boiler plate. My experience adding a protocol to Qt was relatively painless.

We said we'd try on a really simple protocol.

IMHO this looks encouraging.

Obviously having a mix of styles isn't great, but something has to be first.

One thing we need to resolve is

QtWaylandServer::org_kde_kwin_keystate
will call init which will call wl_global_create

Global::Private will call wl_global_create

Calling that twice is wrong.

In D22362#493722, @davidedmundson wrote:

As some background:

At a recent Plasma sprint, one comment was that adding protocols to KWayland was an absolute pain involving writing a tonne of boiler plate. My experience adding a protocol to Qt was relatively painless.

We said we'd try on a really simple protocol.

IMHO this looks encouraging.

Obviously having a mix of styles isn't great, but something has to be first.

---

One thing we need to resolve is

QtWaylandServer::org_kde_kwin_keystate

will call init which will call wl_global_create

Global::Private will call wl_global_create

Calling that twice is wrong.

I was looking into this and realised it's not really easy to solve. The reason is that KWayland::Server::Global was designed to be able to expose the wl_global as if it was useful, but this is something you never do. The wl_global is private on the generated code so we can't access it. I'll provide a workaround for now to keep the conversation alive, but I'd say at least operator wl_global*(); should get deprecated.