Silencing Empowerment, Allowing Bigotry: Auditing the Moderation of Hate Speech on Twitch