refactor(web): drop useDelayed sync-timeout hack; fix test timing instead

The synchronous pre-timeout in the implementation was a band-aid for a
missing flushSync() in the third test. $effect callbacks are scheduled
as microtasks and don't run until flushed — the test set source=true
before the initial effect had a chance to register the setTimeout, so
advanceTimersByTime fired nothing.

Proper fix: call flushSync() after $effect.root so the initial effect
runs synchronously, then advance fake timers. Removed the duplicate
setTimeout branch from useDelayed — one code path, no race.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

web/src/lib/utils/useDelayed.svelte.test.ts

@@ -61,6 +61,7 @@ describe('useDelayed', () => {
       const cleanup = $effect.root(() => {
         delayed = useDelayed(() => source, 100);
       });
+      flushSync(); // run the initial $effect so the setTimeout is scheduled
 
       vi.advanceTimersByTime(200);
       expect(delayed.value).toBe(true);

web/src/lib/utils/useDelayed.svelte.ts

@@ -11,15 +11,6 @@ export function useDelayed(
   let delayed = $state(false);
   let timeout: ReturnType<typeof setTimeout> | null = null;
 
-  // Schedule the initial timeout synchronously so tests that advance fake
-  // timers before the first reactive flush still see the correct value.
-  if (source()) {
-    timeout = setTimeout(() => {
-      delayed = true;
-      timeout = null;
-    }, delayMs);
-  }
-
   $effect(() => {
     const v = source();
     if (timeout) {