Fix window offsetting with focus stealing prevention

[tarek-y-ismail]

Closes #3695

Adds occlusion checking when placing down new windows. The checks continue with an increasingly larger radius until no positions around the window fit in the active output or no possible window rectangle overlaps with the active output. The checks are also capped to 16 "iterations" so they don't go on forever, just in case.

Here's how it looks like now:

Screencast.from.2025-01-21.17-58-52.mp4

How it works:

• We repeatedly subtract the rectangles of existing windows from our candidate window rectangle. If after iterating through all windows (or the current window, as an early exit) we still have any visible area of the rectangle, then that placement is used.
• We further filter placements by checking if at least a 50x50 chunk of the app is visible to prevent thin lines that could be hard to notice

[tarek-y-ismail]

There's a bug where a 1280x1024 fullscreen window is returned from for_each_window_in_workspace. It's not clear to me what this window is. My guess is either the background or something leftover from the spinner. Will investigate next week.

[AlanGriffiths]

There's a bug where a 1280x1024 fullscreen window is returned from for_each_window_in_workspace. It's not clear to me what this window is. My guess is either the background or something leftover from the spinner. Will investigate next week.

Check which layer it is in. You're likely seeing the background

[Saviq]

This looks like a piece of code that could use and would yield nicely to some unit testing?

[Saviq]

Focus stealing prevention shouldn't be a factor when placing new windows, why wouldn't you just use the same in both cases?

[tarek-y-ismail]

This looks like a piece of code that could use and would yield nicely to some unit testing?

Indeed, just wanted your opinions on the base idea before investing more time into it.

Focus stealing prevention shouldn't be a factor when placing new windows, why wouldn't you just use the same in both cases?

Well. For one: thou shalt not change established behavior. But in all seriousness, this much work isn't really needed when focus stealing prevention is off, you just place the window somewhere according to the old logic and it's going to be on top anyway. If you deem it suitable enough for FSP=off, then I'm okay with that