fix(anilist): paginate large libraries to avoid CPU budget kills (#184)

* fix(anilist): paginate large libraries to avoid CPU budget kills

Chunk GraphQL fetches and defer merge, sort, and cover work to update()
ticks so 1000+ anime and manga entries load without blocking callbacks.

* fix(anilist): handle OAuth callback immediately and simplify service helpers

Process browser login tokens synchronously from runStream and file poll, keep a 500ms update interval during OAuth, and avoid duplicate OAuth result emissions from Python.

anilist/widget.luau

@@ -76,9 +76,9 @@ local function render()
   barWidget.setGlyph(glyph)
   barWidget.setGlyphColor(if open then "primary" else "on_surface")
 
-  if snapshot.loading then
+  if snapshot.oauthLoading or snapshot.loading then
     barWidget.setText("")
-    barWidget.setTooltip(tr("tooltip_loading"))
+    barWidget.setTooltip(if snapshot.oauthLoading then tr("tooltip_oauth") else tr("tooltip_loading"))
     return
   end