Return correct hit-test values for title bar buttons on Windows

[Nerixyz]

Description

This Pr returns the correct hit-test values for the title bar buttons (minimize, maximize/restore, and close). This will allow Windows 11 snap layouts to work with Chatterino (and possibly other OS features).

I adapted the implementation from https://github.com/desktop-app/lib_ui/blob/44161f183c255dc0dac7ebe9558a1ca48f5e5258/ui/platform/win/ui_window_win.cpp#L755-L810.

• In WM_NCMOUSE{MOVE, HOVER}, we simulate move events for the title bar buttons, as Qt doesn't generate those.
• In WM_NCMOUSELEAVE (and the first WM_MOUSEMOVE after a move in the non-client area), we clear/leave all hovered buttons.
• In WM_NCLBUTTON{UP, DOWN}, we simulate a click on a button.

From my testing, it seems that we don't need to handle WM_NC{L, M, R, X}BUTTONDBLCLK and WM_NC{M, R, X}BUTTON{DOWN, UP}.
For some future PR (maybe): QWidget::nativeEvent always returns false and does nothing else - we don't need to call it.

When testing, please test as many native gestures/tricks/…, like double-clicking the title bar, as you can.

Closes #4991.
Addresses #1967 (as the window can be dragged in the top left corner now - not 100% sure why).

TODO

• Explain this in more detail
• Add comments for the messages
• Test on multiple monitors
• Test on high-dpi
• Handle default window frames
• Check what the tooltips are about Seems like they're from the OS
• Test on Windows 11 (I can't do that 😭)

_{As for high-dpi, this needs to be changed if #4868 ever lands.}

[pajlada]

I'm not a huge fan of Q_ASSERT*, can you confirm Q_ASSERT_X doesn't fire in our CI-built windows artefacts?
I've confirmed Q_ASSERT* works as expected on my machine on Linux by just adding a Q_ASSERT at the top of main.cpp

Once you've confirmed it works, I'm fine to leave it here and would be more comfortable using it for future GUI code

I will test this on Windows11 now

[pajlada]

Works as expected on Windows 11, thank you!