Add comprehensive VoyageAI embedding models with shared embedding space across different model variants. Implement model merging capability to combine data from multiple sources, enabling seamless integration of VoyageAI models alongside existing providers.

Key improvements:
- New VoyageAI models file with 7 embedding variants (large, standard, lite, nano, multimodal, code, finance, law)
- Flexible embedding dimensions (256/512/1024/2048) with Matryoshka support
- Enhanced GetOpenRouterModelsResponse with merge() method for combining model data
- Cleanup pricing fields by replacing "0" strings with null values
- Bump version to 0.8.0 and update dependencies

This expands the model ecosystem to include state-of-the-art embedding models optimized for retrieval, multimodal understanding, and domain-specific tasks.