[pdftex] pdflatex failures on big-endian architectures (s390x / ppc64)

[Karl Berry]

Hi Hilmar - thanks for the report, as always. But ... yuck :).

    The PDFs all _look_ OK in any PDF viewer, it's just the text extraction
    that fails.

Failure of BigEndian + LaTeX + text extraction. Yikes. Doesn't look easy
to work on. Can't easily run things to see if it happened between 2020
and 2021, or 2021 and 2022. Or whether it might be due to changes in
LaTeX.  I don't have TeX, let alone LaTeX, installations of every (or
any) vintage on BigEndian. Setting that up would take quite some time.

Thanh, do you have any ideas how to reduce this to plain/primitive TeX?
When it comes to text extraction, I am not sure what's required.
(Or ideas about what the problem might be, of course.)

Looking at the pdftexdir/ChangeLog between TL'20 and TL'22, nothing
jumps out at me as likely to be responsible :(.  --thanks, karl.