Not always about you: Prioritizing community needs when developing endangered language technology