use union all when building trait query

union all is more performant as it doesn't need to filter out
duplicates. we should not have duplicates as the primary key
conditions already ensure that.

Change-Id: Ic25f7086b33a98ca77a669732a65be5a95b7f0db

ceilometer/event/storage/impl_sqlalchemy.py

@@ -282,7 +282,7 @@ class Connection(base.Connection):
                     sa.cast(sa.null(), sa.String(255)))
                 .filter(sa.exists().where(
                     models.TraitDatetime.event_id == query.subquery().c.id))
-            ).union(
+            ).union_all(
                 session.query(
                     models.TraitInt.event_id,
                     models.TraitInt.key, sa.null(),