help2man used characters in the range 0x80,...,0x84 to mark special
sections/characters during its processing, but those bytes where also
occurring in other utf-8 characters breaking the output.  For instance
the character '₁' ( a subscript 1), is encoded as "0xE2 0x82 0x81" in
utf-8.

* tools/help2man: Tell perl that input and output should be assumed to
be utf-8.  Also use "private-use codepoints" for those special
characters to avoid any future conflict.