Best-effort: locate the entviz's outer #808080 frame in an arbitrary image
(§6.3 stage 1). Probes three vertical scan-lines (x = 25% / 50% / 75% of the
width); at each, it descends to every frame-gray run and tries to validate a
rectangle there, skipping runs that aren't a real frame (so surrounding
screenshot chrome is tolerated). Returns the first validated frame, or null if
none is found (→ the caller reports unknown and lets the human eye decide).
Best-effort: locate the entviz's outer #808080 frame in an arbitrary image (§6.3 stage 1). Probes three vertical scan-lines (x = 25% / 50% / 75% of the width); at each, it descends to every frame-gray run and tries to validate a rectangle there, skipping runs that aren't a real frame (so surrounding screenshot chrome is tolerated). Returns the first validated frame, or null if none is found (→ the caller reports
unknownand lets the human eye decide).