openai/gpt-chat-latest → A (3:2)

Side A makes substantive architectural improvements by introducing a settlement worker that batches vote persistence and ranking recomputation, adds cached ranking reads (`ranked_items_cached`) to avoid recomputing on every HTTP request, switches ranking access from write to read locks, and removes the obsolete demo-counter feature. Side B consistently migrates the DSL to a new explanation-first vote syntax and updates the parser, UI emission, documentation, and tests, but much of the patch is a project-wide syntax conversion rather than adding new runtime capability.