feat: support sending voice messages (SILK/MP3/OGG)

[lemoncat7]

Summary

Adds support for sending voice messages to WeChat, completing the round-trip for voice media.

What was added

• sendVoiceMessageWeixin() in src/messaging/send.ts

  • Sends voice via CDN upload + MessageItemType.VOICE
  • Supports encode types: 6=SILK, 7=MP3, 8=OGG-SPEEX
  • Optional playtimeMs and encodeType params
  • Optional text caption sent as a separate TEXT item first

• uploadVoiceToWeixin() in src/cdn/upload.ts

  • Uploads audio files with media_type=4 (VOICE) to WeChat CDN
  • Reuses the existing uploadMediaToCdn() pipeline (AES-128-ECB encryption)

• sendWeixinMediaFile() now routes audio/* MIME types to the voice path

  • audio/mp3 → encode_type=7
  • audio/ogg → encode_type=8
  • Default → encode_type=6 (SILK)

API surface

// Direct call with pre-uploaded audio
sendVoiceMessageWeixin({
  to: "wxid_xxx",
  text: "Optional caption",
  uploaded: { filekey, downloadEncryptedQueryParam, aeskey, fileSize, fileSizeCiphertext },
  encodeType: 6, // SILK
  playtimeMs: 5000,
  opts: { baseUrl, token, contextToken }
})

// Via sendWeixinMediaFile (auto-routes by MIME type)
sendWeixinMediaFile({ filePath: "/path/to/audio.mp3", to: "wxid_xxx", text: "...", opts, cdnBaseUrl })

Note

Receiving voice was already implemented (silk-transcode.ts + VoiceItem). This PR adds the send direction, making voice a full two-way media type.

---

Closes #feature/voice-send