openai/gpt-chat-latest → A (4:1)

Side A fixes importer behavior by explicitly skipping stickied/pinned Reddit posts in `parse_children`, adds a dedicated `child_is_pinned` helper checking both `stickied` and `pinned` flags, and includes a regression test covering both cases. Side B mainly removes a guard that is asserted to be unreachable because earlier validation already rejects zero ratios, updating tests to reflect existing behavior rather than adding new functionality or fixing a user-visible issue.