feat(meditate): dreams spoken: a microphone in the dream's editor, and Today's Speak it

The editor dictates where the cursor stands with halkyon_app's Dictation,
showing it listens and the words heard, listening again after each pause
until stopped or three silent listens, in the app's language. No
microphone where the device has no recogniser (Linux). On Android it asks
for recognition on the device; the privacy page says what each platform
does. Android's, iOS's and macOS's microphone declarations as Learn's.
ADR 0038, D15.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>

apps/meditate/halkyon_meditate_flutter/android/app/src/main/AndroidManifest.xml

@@ -3,6 +3,10 @@
     <!-- The app talks to its instance: release builds need this too (Flutter's
          template grants it to debug and profile builds only). -->
     <uses-permission android:name="android.permission.INTERNET"/>
+    <!-- A dream spoken into the journal (ADR 0038, D15): the microphone, asked
+         for once the person taps it, and the device's speech recogniser, which
+         Android 11 and later only shows the app when it is declared below. -->
+    <uses-permission android:name="android.permission.RECORD_AUDIO"/>
     <!-- No backup and no transfer to another phone (ADR 0038, D7, as
          Train's ADR 0030, D53): what is sealed comes back from the
          instance, and a backup would hold the device's opened copy. Through Android 11 allowBackup and
@@ -66,5 +70,8 @@
             <action android:name="android.intent.action.PROCESS_TEXT"/>
             <data android:mimeType="text/plain"/>
         </intent>
+        <intent>
+            <action android:name="android.speech.RecognitionService"/>
+        </intent>
     </queries>
 </manifest>