feat(workspace): dictation via /diktat (M4)

/diktat (only in an empty bullet, only with voice input available)
listens right away; the bullet becomes one button with the live wave
across it and a translucent "Stopp" on top. A tap ends it and the
recognised text lands in the bullet as plain text. Uses the configured
speech recognition (own Whisper server or the browser); Esc cancels.

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

frontend/src/features/help/HelpView.tsx

@@ -162,6 +162,7 @@ export const HelpView: React.FC<HelpViewProps> = ({ onBack }) => {
           'bild',
           'audio',
           'sprachnotiz',
+          'diktat',
           'image',
           'foto',
           'auto-formatting',
@@ -205,6 +206,10 @@ export const HelpView: React.FC<HelpViewProps> = ({ onBack }) => {
                   <span className="text-accent font-bold">🎙️</span>
                   <span>/audio (Sprachnotiz aufnehmen: grüner Knopf startet, roter speichert; oder Audiodatei hochladen)</span>
                 </div>
+                <div className="p-1.5 bg-panel rounded border border-border flex items-center gap-1.5 text-main col-span-2">
+                  <span className="text-accent font-bold">🗣️</span>
+                  <span>/diktat (sprechen statt tippen: startet sofort, Tipp auf „Stopp“ setzt den Text in den Punkt)</span>
+                </div>
               </div>
             </div>
             <div className="pt-2 border-t border-border">