Leave the character encoding from the locale, but prefer the familiar
numeric encoding.

This fixes a bug where the scan script did not match the numeric
parsing, and the man page or help message did not match the way
options were parsed.

Thanks to Daniel for the suggestion, information and providing
an initial patch. He encountered these problems using LANG=de_DE.UTF-8
which I also used in my test.

We also considered fully honouring the encoding, but decided that
would be a bigger change of scan scripts, man page etc. and wouldn't
necessarily match user expectations either.

Requested-by: Daniel Holbach <daniel@holba.ch>