pdf_production.py produces incorrect results

[mzilinec]

Hello, while experimenting with PERO I have noticed a few problems with pero_ocr.document_ocr.pdf_production, namely:

• in my case, resulting PDF lines from the OCR were consistently offset (one line below the original line in the PDF) - example below
• occasionally, fpdf would produce a page break when text is too close to the border, causing the output to be invalid,
• the script crashes due to fonts being moved to a different location in newer linux distributions

This can be easily fixed by:

• replacing bottom with bottom - line.heights[0] + line.heights[1]
• calling pdf.set_auto_page_break(False)
• making the font into a default parameter

If you want, I can send a PR.

Example output: