Revisiting Binary Code Similarity Analysis Using Interpretable Feature Engineering and Lessons Learned