Few minor performance improvements

This is unnecessary and could even harm performance (depending on your system.) It was used in AlbumView and ThumbnailStrip. Also for some reason displayMarginBeginning/displayMarginEnd was set for the thumbnail strip but I don't know why, so I removed that too.

For example, on my 4K screen at 125% scaling the album view can create a staggering >2k delegates with our current cacheBuffer sizing! To put that into perspective, using the default cacheBuffer size results in only ~900 (which is still a lot IMO but at least Qt is deciding that.) At 31x19 rows and columns, only ~589 can be shown at once anyway which makes a large amount of these delegates redundant.

They also take up significant memory - at least hundreds of megabytes - and it also adds a few seconds to our initialization time. Naturally this depends on the size of your screen and what you're viewing, it's all relative though. I didn't notice scrolling performance improving or getting worse when going back to the default cacheBuffer size.

---

I set the background to null instead of creating an empty Item (this has no performance impact AFAIK it gets optimized out.)

I also removed unused selection rectangles from the scene graph entirely because transparency alone doesn't do that. Then I changed the corner radius to use our standard Kirigami cornerRadius unit instead of "2".