Don't munge reply-to if one already exists -- feature reques

As you all know, a few (loud) users still seem to think reply-to header
munging is a good idea, and thus I and many others feel forced to enable it
on certain lists. This causes problems every time people would like a reply
off-list, since people tend to reply to the list by mistake, obeying the
munged reply-to.

Turning off first_strip_reply_to looks like it would help at first glance,
but it still requires users to edit the headers manually to remove the
added list address, and many forget this.

A simple solution would be to simply allow leaving any existing
reply-to-headers alone, and only add one if none are already defined. This
way, the munging will act as a default, allowing the sender to override it
by adding an explicit one.

The change should be fairly simple, and something similar to the attached
diff should suffice. Naturally, the setting "leave_existing_reply_to_alone"
would also have to be added and documented. The diff is just to illustrate
the idea (although I can create a complete diff if you wish).

[http://sourceforge.net/tracker/index.php?func=detail&aid=1707731&group_id=103&atid=350103]