hermes - 💡(How to fix) Fix fix(gateway): Telegram sends voice + text as separate messages on auto-TTS [1 pull requests]

Bug description

When auto-TTS fires on Telegram (voice input with voice_only mode active), the gateway sends two separate messages:

1. A native voice bubble (auto-TTS audio)
2. A text message with the response content

Expected behaviour: one message — a voice bubble with the text as caption (similar to how Discord correctly sends only the voice).

Steps to reproduce

1. Send a voice message to the Telegram bot
2. Observe that both a voice bubble and a text message appear
3. Compare with Discord where only a voice bubble appears

Platform

• Telegram bot adapter
• Auto-TTS enabled (voice.auto_tts: true in config)

Additional context

This is similar to Signal issue #8250 (fix(signal): preserve native TTS format and send single captioned audio reply). The Signal adapter has the same issue. The Telegram send_voice method supports captions, so the fix should make the auto-TTS path send the text as a caption on the voice message instead of as a separate text message.

Relevant code in base.py around line 2816-2844 — the auto-TTS path sends voice via play_tts() and then unconditionally sends text via _send_with_retry() regardless of whether voice was already sent.