add sorter to itermatch, and optimize the piss out of everything else to use it.

upshot of this, sorting occurs per repo.itermatch return- so for "prefer highest", we've got a set of iterables in sorted order.
Enter util.iterables.iter_sort; func, given a sorting func, maintains iterable order, yielding the cumulative highest.

End result is the same as if you just used a sort on a list, but it pulls only as needed- since for any repo that uses metadata
based visibility filters, this means their *always* will be disk activity, thus converting matches to a list and sorting accesses
packages that may not even be needed.  Iteratively sort it in combination with iter_caching, and you access only what's needed.
For system testing, knocked sys time down by 66%.

1 files changed, 1 insertions, 1 deletions

diff --git a/bin/utilities/pquery b/bin/utilities/pquery
index bea044ab0..baa125943 100755
--- a/bin/utilities/pquery
+++ b/bin/utilities/pquery
@@ -131,7 +131,7 @@ if __name__ == "__main__":
 		atoms = packages.OrRestriction(*atoms)
 	for a in atoms:
 		for repo in repos:
-			for pkg in results_filter(repo.itermatch(a)):
+			for pkg in results_filter(repo.itermatch(a, sorter=pre_curry(sorted, reverse=True))):
 				try:	print str(atomstr(pkg))
 				except AttributeError:
 					print str(pkg)