FAIRTOPIA: Envisioning Multi-Agent Guardianship for Disrupting Unfair AI Pipelines