In our test, fictional employees with male names got significantly more than women in bonuses based on retention, but about the same when it came to rewarding potential.