Regenerate proto/gui_gtk4.pro after gui_gtk4_x11.c was merged into
gui_gtk4.c.  Also add gui_gtk4.pro to ALL_GUI_PRO and the dependency
rules in the Makefile so `make proto` picks it up.

The GTK_MENU_SHELL() stub was defined with no parameters, but callers
in getchar.c pass a widget pointer (matching the original GTK macro
signature).  This caused a compilation error.

gui_get_x11_windis() was declared as void but gui.c expects an int
return value (compared against OK).  Return FAIL since X11 window/
display are not applicable in GTK4.

GTK4 does not have GtkMenuShell.  Instead of providing stub functions
for GTK_MENU_SHELL() and gtk_menu_shell_select_first(), guard the
F10 menu activation code in getchar.c with !defined(USE_GTK4) and
remove the stubs from gui_gtk4.c.

gtk_widget_show() is deprecated in GTK4.  Use gtk_widget_set_visible()
with TRUE instead.

- Replace gtk_widget_show/hide() with gtk_widget_set_visible()
- Replace gtk_style_context_add_provider() with
  gtk_style_context_add_provider_for_display()
- Guard gtk_widget_get_style_context() with #ifndef USE_GTK4

GTK4 does not ship gdkkeysyms-compat.h, so legacy symbols like
GDK_Shift_L and GDK_MOD1_MASK are not available.  Use GDK_KEY_Shift_L
and GDK_ALT_MASK under USE_GTK4.

GTK4's GtkIMContext does not support synthesizing key events for IM
activation, so 'imactivatekey' cannot work.  Return FALSE for non-empty
values so that invalid settings are rejected instead of silently ignored.

Calling gtk_init() before fork() breaks the display connection in the
child process, making gvim fail to start without --nofork.  Move the
call to gui_mch_init_check() which is called after fork(), matching
the GTK3 pattern of deferring gtk_init_check() until after fork.

…t signature

gui_mch_newfont() was calling gui_set_shellsize() which called
gui_mch_newfont() back, causing infinite recursion and a crash.
Use gui_resize_shell() instead to recalculate Rows/Columns from the
current window size without re-entering gui_set_shellsize().

Also add a FIXME comment to gui.c where gui_mch_newfont() is called
even when the font hasn't changed (e.g. just setting guioptions=k).

Without this, WM_CLASS is empty on X11, which prevents window managers
like dwm from identifying gvim windows for placement rules.

Port the ligature rendering logic from gui_gtk_x11.c (GTK3) to GTK4.
Previously ligature characters were rendered individually through the
fast ASCII glyph cache path, ignoring the ligatures_map entirely.

Split gui_gtk_draw_string into a wrapper that segments the string into
ASCII and ligature/UTF-8 parts, and gui_gtk_draw_string_ext that does
the actual drawing with proper Pango shaping via pango_shape_full().
Also port helper functions for cluster-based glyph width calculation,
combining character handling, and guifontwide support.

Two issues prevented mouse events from reaching Vim:

1. gui.formwin (GtkForm overlay for scrollbars) was layered on top of
   gui.drawarea via GtkOverlay, intercepting all mouse events.
   Fix: set can_target=FALSE on formwin so events pass through.

2. Drag detection used `if (mouse_pressed_button)` but MOUSE_LEFT is
   0x00, so left-button drags were never detected.
   Fix: use -1 as "no button pressed" sentinel and check >= 0.

gui_mch_flush() only called gdk_display_flush() which flushes the
display buffer but does not notify GTK that the offscreen surface
content has changed. Add gtk_widget_queue_draw() so the draw_event
callback is triggered to paint the updated surface to the widget.

When focus moved to another widget (e.g. menubar, toolbar), the
drawing area did not regain focus on mouse re-enter, leaving the
cursor in its unfocused (hollow) shape. Add gtk_widget_grab_focus()
in enter_notify_event, matching the GTK3 behavior.

When the drawing area was resized (e.g. after :vsp), the old surface
content was copied to the new surface. If Vim's redraw didn't cover
all areas, stale content like the intro screen text remained as ghost
artifacts. Fix by filling the new surface with the background color
instead of copying old content, since gui_resize_shell() triggers a
full redraw anyway. Also remove duplicate surface resizing logic from
draw_event, centralizing it in drawarea_resize_cb.

WM_CLASS is an X11 property that GTK4 no longer sets directly.
The correct way to associate windows with the application in GTK4
is through the StartupWMClass field in the desktop entry file.

When 'guioptions' includes "c" and 'confirm' is set, :q on a modified
buffer sets exiting=TRUE before calling do_dialog().  The GTK4-specific
check for "exiting" in gui_mch_wait_for_chars() caused it to bail out
immediately without waiting for key input, resulting in an infinite
loop.  Remove the "exiting" check to match the GTK3 implementation.

The button label strings parsed from the "&Yes\n&No\n&Cancel" format
were stored as pointers into a buffer that was freed before
gtk_alert_dialog_set_buttons() copied them.  Defer vim_free(buf) and
CONVERT_TO_UTF8_FREE() until after the dialog is done.

Implement gui_mch_menu_grey() for GTK4 using g_simple_action_set_enabled()
to grey out disabled menu items (e.g. Copy/Paste when unavailable).

Store the GAction name in menu->label for later lookup.

Also fix the popover menu not closing after selecting an item.
GTK4's GtkPopoverMenuBar marks popovers as not-visible but Vim's
custom main loop does not process the rendering update.
Use gtk_widget_unrealize() to force the popover surface to close.

GTK4 does not provide a window position API (removed by design for
Wayland compatibility). Return FAIL instead of reporting 0,0 which
is misleading.

List monospace font families from Pango context for :set guifont=<TAB>
completion, matching the GTK3 implementation.

Use GtkPrintOperation with Pango/Cairo rendering instead of
PostScript + lpr. Guarded by FEAT_GUI_GTK_PRINT ifdef.