Add support to Video Generation Models.

[KaifAhmad1]

🚀 Feature Request: Add Support for Video Generation Models

Please consider adding support for the following video generation models, which are currently not supported by lmdeploy:

• CogVideoX-5b by THUDM
• Mochi-1 (Preview) by Genmo
• Allegro by Rhymes AI
• LTX-Video by Lightricks

These models represent the latest advancements in text-to-video generation, and their support would significantly enhance lmdeploy's capabilities in the multimodal AI space.

💡 Additional Context

These models are gaining popularity in generative AI workflows for creating high-quality video content from text prompts. Integration with lmdeploy would enable more efficient inference and unlock a range of creative and industrial use cases.