Steering Large Language Models for Machine Translation with Finetuning and In-Context Learning