Initial support for UTF-8 terminals. At this time, just performs charset

translation from CP437 -> UTF-8 and only when using high-level text output
functions - so the output from native/DOS doors and scripts that use low-level
output methods won't be translated.
UTF-8 is auto-detected (only) at this time during answer (before login) by
sending a ZWNBSP and checking (via ANSI) if the cursor position moved - an
idea "borrowed" from Ozz Nixon (aka SqZ).

This commit requires the new "encode" library (src/encode), so you may need
to perform a cvs checkout to get that module!