The xdg-shell protocol disallows buffer commits to a wl_surface with an
xdg_surface based role before we ack the first configure event. Failing
to adhere to this requirement will get our client disconnected by the
compositor with a protocol error.

Flush a window_surface to a Wayland surface by creating a wl_shm buffer
matching the window size, copying the whole window contents to that
buffer and attaching it to the corresponding Wayland surface.

In our setup with a dedicated event dispatch thread, libwayland ensures
that object proxies associated with an event handler remain valid (or
NULL) while the handler is executing. However, no such guarantees are
given for the proxy user data. It is thus possible for the user data to
become invalid (e.g., its memory freed from a different thread) right
after the event handler is entered.

This is an issue for wayland_surface associated proxies since they may
receive unsolicited events from the compositor at any time (e.g.,
xdg_surface.configure), even while we are destroying the
wayland_surface.

To avoid the problem, we introduce a lock that protects access
to xdg_surface user data and ensures that the associated wayland_surface
remains valid for the duration of the handler.
Co-authored-by: Rémi Bernon <rbernon@codeweavers.com>

Use the xdg-shell Wayland protocol to create Wayland xdg_surface
xdg_toplevel surfaces for non-child windows.

Create and manage an internal driver data structure for each non-desktop,
non-message window.

Limit it to updating user memory and pitch.

Instead of using wined3d_texture_update_desc(). This is safe, because:

* ddraw never exposes wined3d textures directly, and always retrieves them
  directly from wined3d when rendering.

* d3d8 and d3d9 (non-extended) will only resize buffers during a reset, and
  resetting is forbidden if the application holds any references to the
  backbuffers. RTVs are also replaced during a reset, so there is no concern
  about retrieving the old RTVs from the device state.

* d3d9ex allows resetting while holding references to the backbuffers, but tests
  (fixed by this patch) show that the backbuffers should in fact be recreated.

* dxgi forbids holding references to back buffers during ResizeBuffers(),
  including indirect references via command lists or device contexts.

This swapchain is never actually exposed, and effectively only exists due to
implementation constraints. When it was introduced, it was necessary or simplest
to create a dxgi swapchain object, but currently that's no longer necessary, and
avoiding that allows some code to be simplified.

Fixes: 5b11157e
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55141

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55101

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55146

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55146

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55146

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55160

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55137

The state of the caption / nonclient area should be only changed when
FLASHW_CAPTION or FLASHW_STOP are used.