Classroom homogeneity and the distribution of student math performance: A country-level fixed-effects analysis