fix(transcription): neutralize legacy items() to dedupe @mention fetch

Tiptap's suggestion items() callback fired a fetch on every keystroke
after `@`, in parallel with the debounced search-input fetch. Its result
was discarded by updateState, so it was pure waste — doubling the load
on /api/persons and confusing the debounce.

Returning [] from items() routes the entire fetch flow through the
search-input -> debounced onSearch path. New test pins @Walter to
exactly one fetch.

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

frontend/src/lib/shared/discussion/PersonMentionEditor.svelte.spec.ts

@@ -195,6 +195,24 @@ describe('PersonMentionEditor — AC-2/3: search input drives fetch', () => {
 		expect(fetchesAfterSearch).toBe(1);
 	});
 
+	it('fires exactly one /api/persons fetch when the user types @Walter (debounced)', async () => {
+		const fetchMock = vi
+			.fn()
+			.mockResolvedValue({ ok: true, json: vi.fn().mockResolvedValue([AUGUSTE]) });
+		vi.stubGlobal('fetch', fetchMock);
+		renderHost();
+
+		await userEvent.type(page.getByRole('textbox'), '@Walter');
+
+		// Wait beyond the 150 ms debounce window so the trailing call has flushed.
+		await new Promise((r) => setTimeout(r, 300));
+
+		const personsFetches = fetchMock.mock.calls.filter(
+			([url]) => typeof url === 'string' && url.startsWith('/api/persons')
+		);
+		expect(personsFetches.length).toBe(1);
+	});
+
 	it('clearing the search input clears the list without firing a fetch', async () => {
 		const fetchMock = vi
 			.fn()