Order by id instead of time because timestamps may be nonmonotonic.

Probably a result of moderator actions, I found that items with a lower
id could have a higher timestamp, enough to offset it by 2,000 ids.

hnarchive.py

@@ -220,7 +220,7 @@ def select_item(id):
     return item
 
 def select_latest_id():
-    cur = sql.execute('SELECT id FROM items ORDER BY time DESC, id ASC LIMIT 1')
+    cur = sql.execute('SELECT id FROM items ORDER BY id DESC LIMIT 1')
     row = cur.fetchone()
     if row is None:
         return None