Leader selection for minimum-time consensus in multi-agent networks