Vertical text rendering broken in PDFTextDevice

Vertical text rendering broken in PDFTextDevice: a task in harbor-swesmith-rl-artifacts: Harbor SWE-Smith 强化学习数据产物 (Harbor dataset). When extracting text from PDFs that contain vertical text (text written top-to-bottom), the text extraction is broken. The characters are positioned incorrectly…

The task

When extracting text from PDFs that contain vertical text (text written top-to-bottom), the text extraction is broken. The characters are positioned incorrectly, resulting in garbled or incorrect output.

Part of keryszhan/harbor-swesmith-rl-artifacts.