openai/gpt-5.2-chat → B (3:2)

Commit B fixes broken OAuth mock servers and Playwright auth helpers, restoring end-to-end login flows. It addresses crashing handlers, incorrect request body reading, redirect semantics, null token handling, query parsing, and improves test selectors—changes that unblock multiple E2E authentication scenarios. Commit A improves URL display consistency and significantly strengthens a specific vote-pool test with full pair coverage and ranking assertions, but its scope is narrower. Restoring the auth test infrastructure has broader impact across the test suite and developer workflow.