Performance improvement in pdf_try_load_font()

Creating an issue for your visibility and tracking. I found that there is a low-hanging-fruit optimization for parsing large PDFs containing thousands of font entries. I filed a bug with MuPDF upstream and they confirmed my findings, but have given no timeline on a fix (which is reasonable, this fix is a "nice to have")

When this change comes out, it might be worth bumping your MuPDF version to pick it up: https://bugs.ghostscript.com/show_bug.cgi?id=709700