If yes, can I simply convert the XGBoost model into a format that golang GBDT library understands, and use it for evaluation?
Is the evaluation of XGBoost and GBDT same?
That should be possible, you just need to translate the model structure and conditions from XGBoost to golang GBDT.
For an example of generalizing XGBoost models look at Treelite