Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of

making pull_up_sublinks() construct a full-blown JoinExpr tree representation
of IN/EXISTS SubLinks that it is able to convert to semi or anti joins.
This makes pull_up_sublinks() a shade more complex, but the gain in semantic
clarity is worth it.  I still have more to do in this area to address the
previously-discussed problems, but this commit in itself fixes at least one
bug in HEAD, as shown by added regression test case.

src/test/regress/sql/join.sql

@@ -495,3 +495,13 @@ select a.unique2, a.ten, b.tenthous, b.unique2, b.hundred
 from tenk1 a left join tenk1 b on a.unique2 = b.tenthous
 where a.unique1 = 42 and
       ((b.unique2 is null and a.ten = 2) or b.hundred = 3);
+
+--
+-- test proper positioning of one-time quals in EXISTS (8.4devel bug)
+--
+prepare foo(bool) as
+  select count(*) from tenk1 a left join tenk1 b
+    on (a.unique2 = b.unique1 and exists
+        (select 1 from tenk1 c where c.thousand = b.unique2 and $1));
+execute foo(true);
+execute foo(false);