[updated patch] malloc per-thread cache ready for review

[DJ Delorie]

I assume I may need to tweak the patch to fit in the absolute latest
copy of everything, if for no other reason than last-minute-merge
conflicts.  I'll shift that text around at that time, unless you want to
see a new patch with that change?

(git has mangled the tcache branch, so I can merge master to it but
taking diffs off it results in the wrong diff, so I have to keep that
tree and master updated together, which inspires me to delay such syncs :-P )