Skip autoscaling test until more reliable

AutoScalingTest has issues which will require rethinking how
this test is implemented, including:
- the optimal ConsumeStopSeconds will depend on the overhead
  of boot time, which varies greatly between bare metal and
  virt-on-virt
- every combination of distro/release/flavor will result in a
  different baseline memory usage, making it difficult to
  determine the ScaleDownThreshold to set

This change skips this test and tracks against the bug:
Partial-Bug: #1257575

Change-Id: I0695a7efacdbeab39c895a97cb45496fcaf97e35
diff --git a/tempest/scenario/orchestration/test_autoscaling.py b/tempest/scenario/orchestration/test_autoscaling.py
index e843793..90ef3a0 100644
--- a/tempest/scenario/orchestration/test_autoscaling.py
+++ b/tempest/scenario/orchestration/test_autoscaling.py
@@ -19,6 +19,7 @@
 from tempest.test import attr
 from tempest.test import call_until_true
 from tempest.test import services
+from tempest.test import skip_because
 
 
 class AutoScalingTest(manager.OrchestrationScenarioTest):
@@ -62,6 +63,7 @@
         if not self.config.orchestration.keypair_name:
             self.set_resource('stack', self.stack)
 
+    @skip_because(bug="1257575")
     @attr(type='slow')
     @services('orchestration', 'compute')
     def test_scale_up_then_down(self):