[wcook04]

Opening a statement on a phone added a Back button that pushed the reading/PDF controls beyond the screen. Let the existing header controls wrap, and add a browser regression for that state.

Correct the README, contributor guide, page footer and generated text description to explain what the PDF comparison actually checks. It skips some page regions and permits text rearrangements, so passing does not certify an exact copy. Use the tested installation command and describe the required result-and-passage entries. The original paper, PDF, Lean project and author credits are unchanged.

Validation:

• Unit tests, source checks, paper checks, TypeScript and the production build pass.
• A fresh clone adds a small second paper, extracts its three Lean declarations and comments, checks the links and builds a working reader. The synthetic fixture is kept out of the published papers.
• Actual browser checks cover the original and second paper at desktop, laptop and 390-pixel phone widths, including PDF display, exact Lean lines, Back/Forward and control bounds. The new overflow assertions fail on the saved original build and pass on this change.
• The local scripted browser journey could not finish launching. Its new regression steps were run through the connected browser instead; no scripted journey pass is claimed.