data: kvKbPerToken column (exact KV-cache cost for hybrid linear-attention models)

data: 107 models - adds Xiaomi MiMo-V2-Flash + ggufDiy column

data: add openWeights column (sync with GitHub)

Create models.csv