mailinfo: assume input is latin-1 on the header as we do for the body

mailinfo: assume input is latin-1 on the header as we do for the body

When the input mbox does not identify what encoding it is in,
and already have RFC2047 stripped away, we cannot tell what
encoding the header text is in.  For body text, when the message
does not say what charset it is in, we fall back to assume
latin-1 input when converting to utf8.  This should be done
consistently to the header as well.

Signed-off-by: Junio C Hamano <junkio@cox.net>