smbd: Fix a connection run-down race condition

smbd: Fix a connection run-down race condition

When we do a server exit with active aio jobs, we need to keep the
aio state active for the helper thread. Right now I don't see another
chance than to leak memory in this case. And, I don't really oversee
how cancelling requests works in this case, but this does fix crashes
seen at a customer site.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=12925

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
(cherry picked from commit 0181fcc4aaa730e3a88ff5d397145332f4013950)