Add Gemini STT, Gemini TTS options, and WhatsApp PTT delivery

Problem

Hermes has WhatsApp and voice-related surfaces, but the end-to-end path for Gemini speech-to-text, configurable Gemini text-to-speech, and native WhatsApp push-to-talk delivery is incomplete.

Fix

Add three coordinated pieces:

1. tools/transcription_tools.py: add Gemini as a first-class STT provider using the Google Generative Language API.
2. tools/tts_tool.py: expose Gemini TTS model, voice, and output codec configuration, including OGG Opus output.
3. gateway/platforms/whatsapp.py: add a send_voice override that can deliver supported audio as WhatsApp PTT and fall back to the generic audio path when needed.

The change is opt-in. Existing STT, TTS, and send_audio behavior is unchanged when Gemini configuration is absent.

Configuration

GEMINI_STT_API_KEY=...
GEMINI_STT_MODEL=gemini-3-flash
GEMINI_TTS_API_KEY=...
GEMINI_TTS_MODEL=gemini-3.1-flash-tts-preview
GEMINI_TTS_VOICE=Kore

Files touched

• tools/transcription_tools.py
• tools/tts_tool.py
• gateway/platforms/whatsapp.py

Verification

• Syntax checked with python3 -m py_compile on the three touched files.
• Suggested follow-up: add fixture-based STT and TTS tests plus a WhatsApp adapter test for send_voice.