Small-sample adjustments for multiple-contrast hypothesis tests of meta-regressions using robust variance estimation