Bound by the Bounty: Collaboratively Shaping Evaluation Processes for Queer AI Harms