Use hashtable for storage

[koekeishiya]

Uses hashtable for storage instead of having separate tracked arrays for windows and borders.
Also fixes bug mentioned in #6

[koekeishiya]

The hashtable is just copied out of yabai code. The rest is kinda janky again, sorry for low quality code -- most changes are kinda rushed and I have not spent too much time looking at how all the code is organized.

Again, feel free to readjust w/e before merging. You probably want to do another code formatting pass to get style that fits with your code.

[FelixKratz]

I think the PR has one major problem. When moving a window to a different space, there is a brief moment when two borders with the same wid exist, because the window is created at the target space first and then destroyed at the origin space. Thats why I used wid and sid checks to find the window border. I am not sure if sticky windows will fall into the same problem.
This leads to the window "loosing" its border on the target space, because it is destroyed just after being created.

So my suggestion would be to use a hash that is combined in some clever way from wid and sid to identify the window. Or the event logic needs to be adapted to capture this eventuality.

[koekeishiya]

I believe the latest commit should adress that problem just fine.

Edit: I did not do any testing with sticky windows, that might be broken.

Edit2: I might have overlooked some technical differences in border_remove and window_remove when doing these changes, by the looks of it. Might be better to scrap this PR and do changes from scratch with the caveats you mentioned in mind from the start.

[koekeishiya]

Don't merge this. I definitely missed something in your logic and screwed something important. Upon moving a window to a different space (and following focus), moving back to the previous space, now all other windows lack their border.

[FelixKratz]

It could also be I destroyed something during merging…

[koekeishiya]

Ok so I am definitely breaking your logic with the changes in this PR, but it seems to work the way we want it to.
I'll let you make the decisions.

I changed the response to EVENT_UNHIDE and EVENT_HIDE to look for existing border and call SLSOrderWindow to manipulate visibility instead of it calling create/destroy. This appears to mitigate the issue I mentioned in my comment above.

Probably smart to test-drive for a while. There is likely more event logic that need adjustments.

[FelixKratz]

Ok so I have cleaned up and refactored this a bit, what do you think? I am happy with it and will merge once you approve my changes.

[koekeishiya]

Looks good to me.