freeze for sbcl-2.6.5

Hi all,

Let's press pause on all the exciting changes for now, while we gear up
for sbcl-2.6.5, probably in the latter half of next week.  Testing would
be much appreciated, particularly of Real World Code, because although
we are completely within our rights to change the implementation of
INTERSECTION and UNION, history (and a current system which rhymes a bit
with Boogle Plights) suggests that it's possible to end up embedding
dependencies on the order of elements in the return values of those
notional set functions.  Reports and bug fixes to other software that
embeds those assumptions, as well as to SBCL itself, are even more
welcome in the next week than they would be in general.

Thanks,

Christophe