Supertonic TTS service.

[o-alexandre-felipe]

feat: Add Supertonic TTS speech service

This PR adds a new speech service for Supertonic TTS, a fast, high-quality offline text-to-speech engine using ONNX models.

Features:

• Zero configuration: Models auto-download from Hugging Face Hub on first use (~250MB)
• Fully offline: After initial download, no internet connection required
• Fast inference: Up to 167x faster than real-time on modern hardware
• Multiple voices: 4 voice styles available (M1, M2 - male, F1, F2 - female)
• Configurable quality: Adjustable denoising steps (speed vs quality tradeoff)

Installation:

    pip install "manim-voiceover[supertonic]"

Usage:

    from manim_voiceover.services.supertonic import SupertonicService

    service = SupertonicService()

    service = SupertonicService(
        voice_style="F1",
        total_step=5,
        speed=1.0,
    )

Dependencies added:

• onnxruntime - ONNX model inference
• soundfile - Audio file I/O
• huggingface-hub - Model downloading

Testing:
Verified in clean Docker container. All dependencies install correctly and TTS generates audio successfully.