Introducing IDEFICS: An Open Reproduction of State-of-the-art Visual Langage Model