The convergence of Unmanned Aerial Vehicles (UAVs) and Federated Learning (FL) is creating a new paradigm for decentralized intelligence. In this setup, UAVs are not merely new clients; their unique capabilities introduce fundamentally new structural roles that change how FL systems are designed and operate. This systematic literature review addresses this evolving landscape by answering three key questions: 1) How do UAVs reshape FL architectures and roles? 2) In which key domains does combining UAVs with FL solve critical problems? 3) What challenges and mitigation strategies are reported for UAV–FL integration? To answer these questions, we synthesized findings from 272 peer-reviewed journal articles published through late 2025. The review offers a practitioner-oriented guide. First, it provides an architectural blueprint based on the different roles of UAVs, allowing for effective system design. It then offers a playbook of real-world applications, helping build a strong business case for new projects by highlighting promising domains (e.g., satellite-backed maritime communications and wearable health analytics), and revealing gaps and underexplored opportunities. To ensure robust deployment, we outline key implementation challenges and provide actionable guide to mitigate each specific challenge. Finally, we highlight the key gaps that remain and propose focused directions for future work. A keyword co-occurrence map is also provided as a supporting visualization, showing how topics cluster and where potential integration opportunities may lie.