Improve realm update performance

[peppy]

Coming from #28497.

Not sure why I left this around during the refactor. This is 100% handled by the DetachedBeatmapStore.

Removing this subscription reduces overheads by a huge amount for users with large beatmap databases. My hypothesis is that subscriptions are more expensive based on the number of results matching. This one matches almost every beatmap so removing it is a large win.

Using this patch to test:

diff --git a/osu.Game/OsuGameBase.cs b/osu.Game/OsuGameBase.cs
index dc13924b4f..11151104fc 100644
--- a/osu.Game/OsuGameBase.cs
+++ b/osu.Game/OsuGameBase.cs
@@ -419,6 +419,12 @@ private void load(ReadableKeyCombinationProvider keyCombinationProvider, Framewo
 
         private void updateLanguage() => CurrentLanguage.Value = LanguageExtensions.GetLanguageFor(frameworkLocale.Value, localisationParameters.Value);
 
+        protected override void Update()
+        {
+            base.Update();
+            realm.Run(r => r.Refresh());
+        }
+
         private void addFilesWarning()
         {
             var realmStore = new RealmFileStore(realm, Storage);

with this database provided here and profiling the overhead of importing a failed score:

Before	After

[peppy]

If we want to improve performance further (without completely pulling realm out), the next step would be to have a manual event flow which can trigger global invalidation on the DetachedBeatmapStore without requiring the realm pathway to be hit. But it may be agreed upon at that point that pulling out realm is in our best interest 🤷. Definite for another day.