How do you optimize a large-scale test suite for execution efficiency?