SaLinA: Sequential Learning of Agents