metadata.metering incorrectly prepended to ceilometer query rule key

[William C. Arnold]

A cfn template with the following:
  CPUAlarmHigh:
    Type: OS::Metering::Alarm
    Properties:
      description: Scale-up if the average CPU > 65% for 1 minute
      meter_name: cpu_util
      statistic: avg
      period: '300'
      evaluation_periods: '1'
      threshold: '65'
      alarm_actions:
      - {"Fn::GetAtt": [WebServerScaleUpPolicy, AlarmUrl]}
      repeat_actions: true
      matching_metadata: {'metadata.user_metadata.groupname': {Ref: 'WSG'}}
      comparison_operator: gt

The ceilometer alarm generated has query:
metadata.metering.metadata.user_metadata.groupname == ascbvmm1m8-WSG-f4g7cietkbdq

This should be
metadata.user_metadata.groupname == ascbvmm1m8-WSG-f4g7cietkbdq

The result is that the statistics query always results in insufficient samples because there are no matches against samples.
e.g.
| 8a034599-422a-451b-9a08-eaa72a384282 | ascbvmm1m8-CPUAlarmHigh-ppnck6lw3tkk | insufficient data | True | True | cpu_util > 65.0 during 1 x 300s | None |

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/129961

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/129961
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=bd3078e67812d51fa38dce17dd1009bad1613ad2
Submitter: Jenkins
Branch: master

commit bd3078e67812d51fa38dce17dd1009bad1613ad2
Author: Bill Arnold <email address hidden>
Date: Tue Oct 21 11:30:42 2014 -0400

    Fix legacy alarm matching_metadata broken by recent fix.

    The prefix was being computed incorrectly because meter_name is now
    being stripped from kwargs. This patch computes the prefix before
    kwargs is modified.

    Change-Id: I858dd430a4ce4be5cda41b3dd7faa76b8fdf2cfe
    Closes-Bug: #1383521