Fix not detecting Type 0 (composite) fonts

Previously the function to gather used fonts would only walk certain
parts of the PDF object tree. This worked for PDFs which had TrueType
fonts embedded, since the tree looks like this:

(parents)
  BaseFont: AAABQG+ArialMT
  ...
  FontDescriptor:
    FontFile2: ...
    FontName: AAABQG+ArialMT
    ...

Some PDFs have composite fonts, which look like this:

(parents)
  BaseFont: MUFUZY+ArialMT
  ...
  DescendantFonts: # AN ARRAY!
    - BaseFont: MUFUZY+ArialMT
      FontDescriptor:
        FontFile2: ...
        FontName: MUFUZY+ArialMT
        ...

In addition, the FontDescriptor is actually an indirect reference,
which means it doesn't (directly) have a "keys" attribute.

This fixes the false positives for detecting unembedded fonts, by
adding support to walk the tree for arrays and indirect objects.

Note, I found this article helpful [1].

[1]: https://www.prepressure.com/pdf/basics/fonts

app/embedded_fonts.py

@@ -2,9 +2,10 @@
 from io import BytesIO
 
 from PyPDF2 import PdfFileReader
+from PyPDF2.generic import IndirectObject
 
 
-def contains_unembedded_fonts(pdf_data):
+def contains_unembedded_fonts(pdf_data):  # noqa: C901 (too complex)
     """
     Code adapted from https://gist.github.com/tiarno/8a2995e70cee42f01e79
 
@@ -30,6 +31,11 @@ def walk(obj, fnt, emb):
 
             for k in obj.keys():
                 walk(obj[k], fnt, emb)
+        elif isinstance(obj, list):
+            for child in obj:
+                walk(child, fnt, emb)
+        elif isinstance(obj, IndirectObject):
+            walk(obj.getObject(), fnt, emb)
 
     pdf = PdfFileReader(pdf_data)
     fonts = set()

tests/test_embedded_fonts.py

@@ -16,10 +16,10 @@
 
 
 @pytest.mark.parametrize(['pdf_file', 'has_unembedded_fonts'], [
-    (BytesIO(blank_with_address), True),  # false positive, I think, or maybe because created through Google sheets?
+    (BytesIO(blank_with_address), False),
     (BytesIO(example_dwp_pdf), False),
     (BytesIO(multi_page_pdf), True),
-    (BytesIO(valid_letter), True),   # false positive, I think, or maybe because created through Google sheets?
+    (BytesIO(valid_letter), False)
 ], ids=['blank_with_address', 'example_dwp_pdf', 'multi_page_pdf', 'valid_letter'])
 def test_contains_unembedded_fonts(pdf_file, has_unembedded_fonts):
     assert bool(contains_unembedded_fonts(pdf_file)) == has_unembedded_fonts