Investigating the Multilingual Calibration Effects of Language Model Instruction-Tuning