openai/gpt-chat-latest → A (4:1)

Side A makes a focused behavioral improvement by changing rank row coloring from list-position-based to score-based min–max normalization within each group, adds a dedicated `score_gradient_t` helper, and includes targeted tests for score mapping, ties, and styling behavior. Side B mostly introduces infrastructure and dependencies for a voting feature (new crates, routing, UI hooks, CSS, typed form holes, redirect support), but much of the core functionality is absent from the patch, making it a broader yet less self-contained contribution.