openai/gpt-chat-latest → A (4:1)

Side A changes the UI logic to color rank rows based on each group's actual score range instead of ordinal position, introducing a dedicated `score_gradient_t` function, updating callers, and adding focused tests for normalization, tied scores, and stability. Side B is largely a storage refactor from deque to list with query-time capping and schema updates; while it changes persistence behavior, it mainly reorganizes data handling and removes write-time trimming rather than delivering a comparably clear user-facing improvement or bug fix.