Expire speed fixes backported to STABLE

[Russ Allbery]

I've committed the various changes I made to CURRENT to speed up expire
(particularly fastrm) to STABLE.  Note that while I've tested this fairly
extensively under CURRENT, I haven't specifically tested the STABLE code;
it's the same code and should work the same, but you may want to watch it
closely at first.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>