replied to. This may be used later for thread loading optimization (no need
to load headers if there are no replies) - but for now, at least there will
be a "Replied" attr value displayed in the message header indicating that
the message was replied to. This attribute was previously only used in the
"mail" base (for personal email/netmail).

Also, I think I fixed what could have been an infinite loop if there was SMB
corruption: if a message's thread_next value pointed back (to an earlier
message, with a lower message number), we could have been caught in an
infinite loop looking for the last message in the thread. This is just a
theoretical problem and never reported, but apparently possible with just
the right kind of corruption of the msgbase header.