Fix stutter on returning to song select when updating large beatmap sets

[SevenSec114]

Addresses the stutter reported in #35129, whose regression introducer was bisected to the per-difficulty realm check added in #34914.

When a beatmap set is updated (e.g. returning from gameplay with a large set, or exiting the editor), the carousel's replace handling previously:

• ran a blocking realm.Run(...FindWithRefresh...Detach()) round trip on the update thread for every difficulty of the set (added in Fix errors on exiting from editor if there are no more beatmaps remaining for the ruleset active in song select #34914 to fix Editor: error popup when changing game mode of beatmap #34826, first surfaced as the regression in Game freezes when returning to song select from results screen #35129)

This is unnecessary for the items that are not the current selected beatmap since the crash is only by the RequestSelection given the invalid beatmap.

[bdach]

• Did you write this yourself? Was there "AI" assistance here?
• Test failures will be relevant (it appears hiding difficulties no longer works - will let you figure out if this kills your optimisation or not).

[SevenSec114]

My bad for the test failure. I ran the tests I thought were affected and they passed, so I relied on the github CI for the rest.

Also this change is the actual performance fix:

-Items.ReplaceRange(previousIndex, 1, [matchingNewBeatmap]);
+Items.ReplaceRange(previousIndex, oldSetBeatmaps.Count, newBeatmaps);

but I forgot to flip the rest logic in HandleItemsChanged. Will fix.

Code is written by myself, I spent a lot to figure out the logic. But before committing I found something like comments in my own language may be hard to understand so I polished a lot.

[bdach]

Unsure why this PR is in draft, will probably not attempt review until it's out of draft.

[SevenSec114]

I'm now still working on it to find if there is any other issues appeared due to the logic changing, because I don't want it happened one more time. Please give me more time🙏

[SevenSec114]

I found myself over-engineered. But I'm not sure if this test should be removed later since the crash is caused by the assertion in osu.Game/Screens/Select/FooterButtonOptions.cs that can be solved in other place...

beatmap = realm.Run(r => r.Find<BeatmapInfo>(workingBeatmap.Value.BeatmapInfo.ID)!.ToLive(realm));