Temporary fix for corrupted characters

Fixes Issue #3374, by falling back on the deprecated in-python fonttools
based technique for doing subsetting.  To be removed once the new
MuPDF-based code is a little more mature, or at least once [1, 2] are
fixed.

[1] pymupdf/PyMuPDF#3470
[2] pymupdf/PyMuPDF#3494

plom/create/mergeAndCodePages.py

@@ -455,7 +455,9 @@ def make_PDF(
     # Subsetting requires https://pypi.org/project/fonttools
     # Note: In theory, this could muck around with fonts from the source
     # (i.e., if they were NOT subsetted).  Does not happen with LaTeX.
-    exam.subset_fonts()
+    # TODO: consider doing this only when *we* placed non-ascii (Issue #3381)
+    # TODO: remove this fallback on a future PyMuPDF (see Issue #3374)
+    exam.subset_fonts(fallback=True)
 
     # Add the deflate option to compress the embedded pngs
     # see https://pymupdf.readthedocs.io/en/latest/document/#Document.save