openai/gpt-chat-latest → B (9:1)

Side B fixes a real correctness bug in `server/src/ranking.rs` by switching the Rank Centrality transition matrix to use degree-based `d_max`, preventing oscillation and producing correct rankings for star topologies. It also adds focused Rust and Clojure regression tests with ranking fixtures, while Side A primarily adds a long design plan and a `RouteContext` wrapper around `ThreadNav` that introduces little functional change.