avoid double prefixes for mimed subjects

[Dk379]

this patch checks subject: line for being mime-encoded
with charset, and if this the case, decodes is for the
purpose of prefix checking.

this avoids replies with subject
encoded( Re: [listname] old subject.. )

to be converted to
[listname] encoded( Re: [listname] old subject...)

ad infinitum.

patch originally developed by vaget at vaget dot org
and beautified by me.

[http://sourceforge.net/tracker/index.php?func=detail&aid=528031&group_id=103&atid=300103]

[Loewis (loewis)]

How does this relate to patch

http://sourceforge.net/tracker/index.php?func=detail&aid=498766&group_id=103&atid=300103

[Mikhail Zabaluev (mzabaluev)]

This patch does the job, although EncWord functionality is
somewhat moot: it assumes that any string is encoded using
at most one non-ASCII charset. Which _might_ be wrong.
Decoding into a Unicode string with regard to charsets would
be cleaner, but OTOH it would require codec support and
hence would bring additional complexity and trigger more
untrampled bugs.

[Barry Warsaw (barry)]

Since this patch is against MM2.0, I'm going to reject it.
Security patches only now for MM2.0.x.

On a related note, CookHeaders.py in MM2.1 should have
enough smarts to avoid adding the prefix to encoded headers,
although it could miss prefixes which contain spaces and
where each word is in a separate encoded chunk. I think in
practice this is unlikely.