This MR should be no-op and doesn't change much, but makes some small refactoring in preparation for an improved window state tracker as designed in https://gitlab.winehq.org/wine/wine/-/merge_requests/6569.
Overall with that full MR not much would be changed either, and instead some existing fields are replaced with a cleaner design, until the end of the MR where state requests are made asynchronous and where the new information is used to decide whether to apply or delay client state updates.
I believe that the whole MR would fix the currently often failing d3d/ddraw tests, and more generally fix race conditions that happen with window state updates. I believe that it should also ultimately make it easier/safer to change the current design, to delay the client state updates a bit more and apply them outside of the driver ProcessEvent calls: tracking the pending and host states accurately and independently of the win32 state will let us decide what change needs to be requested to make them match, at any moment.
The latter would fix some existing bug where applications stop responding as they start a new message loop in response to SC_RESTORE commands, and as we do not process X11 events recursively.