Check PDF metadata first, then rebuild from extracted styles

When asked to edit a PDF, read its metadata with pypdf (PdfReader(path).metadata) before searching disk for a source file. /Author: python-docx plus /Producer: LibreOffice showed it came from a generated docx, which may no longer exist. If no source turns up, PyMuPDF's page.get_text('dict') gives exact font, size, color and x0 for each span, enough to recreate the look in python-docx and convert with soffice --headless --convert-to pdf. Check the page count after each conversion: the rebuild can overflow to 2 pages even with the same specs, so tighten line spacing and margins until it fits.