#15109: revert '%'->'format' changes in 4b105d328fe7 to fix regression.

With '%', non-ascii worked because the '%' automatically got promoted to
unicode.  With format that doesn't happen, which led to encoding errors.  This
fix goes back to using %, and adds a test to make sure non-ascii string values
work in iterdump.
diff --git a/Lib/sqlite3/dump.py b/Lib/sqlite3/dump.py
index de9c368..e5c5ef2 100644
--- a/Lib/sqlite3/dump.py
+++ b/Lib/sqlite3/dump.py
@@ -43,7 +43,7 @@
         #        qtable,
         #        sql.replace("''")))
         else:
-            yield('{0};'.format(sql))
+            yield('%s;' % sql)
 
         # Build the insert statement for each row of the current table
         table_name_ident = table_name.replace('"', '""')
@@ -54,7 +54,7 @@
             ",".join("""'||quote("{0}")||'""".format(col.replace('"', '""')) for col in column_names))
         query_res = cu.execute(q)
         for row in query_res:
-            yield("{0};".format(row[0]))
+            yield("%s;" % row[0])
 
     # Now when the type is 'index', 'trigger', or 'view'
     q = """
@@ -65,6 +65,6 @@
         """
     schema_res = cu.execute(q)
     for name, type, sql in schema_res.fetchall():
-        yield('{0};'.format(sql))
+        yield('%s;' % sql)
 
     yield('COMMIT;')