Add support for local model path in from_pretrained methods

[mattoofahad]

Add support for local model path in from_pretrained methods

Enhanced the from_pretrained class methods in both ChatterboxMultilingualTTS, ChatterboxTTS, ChatterboxVC and ChatterboxTurboTTS to accept an optional model_snapshot_path parameter via kwargs.

Changes

• Modified from_pretrained methods to accept **kwargs
• Added logic to check for model_snapshot_path parameter
• If model_snapshot_path exists and is valid, the method uses the local path directly, skipping HuggingFace snapshot download
• If model_snapshot_path is not provided or doesn't exist, falls back to the default behavior of downloading from HuggingFace

Benefits

• Allows users to load models from local directories without re-downloading
• Reduces bandwidth usage and initialization time for users with cached models
• Maintains backward compatibility - existing code without model_snapshot_path works unchanged
• Graceful fallback ensures robustness

Usage Example

# Use local model path (skips download)
model = ChatterboxMultilingualTTS.from_pretrained(
    device=device, 
    model_snapshot_path="/path/to/local/model"
)

# Default behavior (downloads from HuggingFace)
model = ChatterboxMultilingualTTS.from_pretrained(device=device)