Liquid AI Releases LFM2-VL: Super-Fast, Open-Weight Vision-Language Models Designed for Low-Latency and Device-Aware Deployment