Theory-informed experimental design for model identifiability: what needs to be measured?