Feature: Add voice message recording and sending in chat

[mjkatgithub]

Goal

Enable users to record audio messages directly in chat and send them as
voice messages.

Scope

• Add microphone recording control in message input (start/pause/stop/cancel).
• Request and handle microphone permissions with clear user feedback.
• Show recording state and duration while capturing audio.
• Generate a preview before sending (playback + delete option).
• Upload recorded audio using existing media upload flow.
• Send uploaded recording as a chat audio message with metadata.
• Render voice messages in timeline with playback controls.
• Handle error states (permission denied, recording failed, upload failed).
• Add/extend unit, integration, and E2E tests.

Branch

feature/voice-message-recording

Acceptance Criteria

• User can start and stop recording from message input.
• Permission flow is handled gracefully (allow/deny).
• User can preview and remove recording before sending.
• Recorded voice message uploads and sends successfully.
• Sent voice message appears in timeline and is playable.
• Recording/upload/send failures show clear error feedback.
• Existing text/image/video flows remain unaffected.

Test Checklist

Unit

• Recording state machine (idle/recording/preview/sending/error).
• Permission-denied and recording-failure handling.
• Voice message payload creation from recorded blob.

Integration

• Recorded blob upload returns media URL/metadata correctly.
• Chat send flow publishes playable voice message event.
• Error mapping from upload/send to UI messages.

E2E

• User records, previews, and sends a voice message in a room.
• Receiver sees and plays the voice message.
• Permission denied path shows actionable feedback.

Notes

• This issue is only for recording in-app via microphone.
• Audio file picker from device storage is out of scope (separate feature).

Protocol/Compatibility Notes (Matrix)

• Send as m.room.message audio event for broad compatibility.

• Include sufficient audio metadata (mimetype, size, duration if known).

• For voice recordings, include voice-message specific markers where supported.

• Always provide fallback rendering as regular audio (m.audio) for clients
  without voice-message UI support.

• Verify behavior against at least one modern Matrix client and one older/basic
  client rendering path.

• If voice-specific fields are unsupported by the receiving client, message must
  still be playable as a normal audio attachment.