fix(voice): browser speech recognition on Android no longer repeats phrases

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

frontend/src/lib/__tests__/voice.join.test.ts

@@ -0,0 +1,21 @@
+import { describe, it, expect } from 'vitest';
+import { joinSpeechResults } from '../voice';
+
+describe('joinSpeechResults (browser speech recognition)', () => {
+  it('Android: growing interim states of one phrase count once', () => {
+    expect(joinSpeechResults(['Test', 'Test', 'Test', 'Test', 'Test', 'Test 1 2 3'])).toBe('Test 1 2 3');
+    expect(joinSpeechResults(['Test', 'Test 1', 'Test 1 2', 'Test 1 2 3'])).toBe('Test 1 2 3');
+  });
+
+  it('desktop: consecutive pieces are joined', () => {
+    expect(joinSpeechResults(['Milch kaufen', ' und Brot'])).toBe('Milch kaufen und Brot');
+  });
+
+  it('Android: a new phrase after a finished one is appended', () => {
+    expect(joinSpeechResults(['Milch', 'Milch kaufen', ' und Brot', ' und Brot holen'])).toBe('Milch kaufen und Brot holen');
+  });
+
+  it('ignores empty results and earlier shorter states', () => {
+    expect(joinSpeechResults(['', 'Test 1 2', 'Test 1', ''])).toBe('Test 1 2');
+  });
+});