Garbled pdftotext output from Type3-font lease PDFs: render pages instead

Some e-signed lease PDFs (NAA/Entrata-style forms) embed Type3 fonts with no ToUnicode map, so pdftotext -layout returns mostly mojibake while only the typed-in fields extract cleanly. Grep the garbled text to find which page numbers mention the target section, then render just those pages with pdftoppm -f N -l N -r 110 -png and read the images. Rendering at about 110 dpi is enough to read form fields and checkboxes without huge images; send pdftoppm stderr to /dev/null, since it emits thousands of Bad bounding box warnings.