pymupdf/PyMuPDF 1.16.5

Fix #381 part 2

on GitHub

On rare occasions, page rendering and text extraction caused interpreter segmentation faults.
When fonts on a page (illegally) were meeting all of the following conditions:

1. a non-UTF8 encodable name
2. not embedded in the PDF
3. using Identity-H / Identity-V ancoding

Then MuPDF reacted with appropriate warning messages (which are intercepted by PyMuPDF).
In prior versions, PyMuPDF did not expect non-UTF8 messages and used a too "optimistic" char to unicode conversion method, which brought down the interpreter.
This has changed in the new version.