Sequence-break markers may be hard to see at real hole density

#141 marks non-consecutive acquisition steps within a square by drawing that path segment as a dashed red line (`#cf222e`, dash `strokeWidth*3 strokeWidth*2`).

Verified working at 24 holes, where it reads clearly. The concern is density: on a square with hundreds of foilholes, adjacent hole centres are only a few pixels apart, so the break segment is a very short dashed stub between two touching circles. In the Numbered and Sparse styles it is further muted, since those segments render at 0.5 opacity against 0.9 for the gradient style.

Worth confirming against a real dense square before deciding whether anything is needed. Possible directions if it does turn out to be illegible:

- Scale the break marker independently of `strokeWidth` so it does not shrink with the path line
- Mark the break on the holes either side (e.g. a small ring or notch) rather than only on the segment between them
- Lift the break-segment opacity in the Numbered and Sparse styles to match the gradient style
- Surface a break count in the square header, so the presence of breaks is discoverable without hunting for them visually

Not a defect in #141 — the logic is correct and confirmed against seeded data with breaks after holes 8 and 16, with no false positives.

No file or test is named. Start by reviewing the sequence-break marker work in #141 and reproducing the seeded dense-square case described here, including Numbered, Sparse, and gradient styles. Done means confirming whether break markers remain legible at real hole density and, if not, agreeing on and implementing one visibility improvement with coverage for the dense case.

Written by the indexing model from the issue text.