Addresses: BP exclude-low-sample-count
To avoid alarm threshold evaluation being inappropriately skewed by
aggregates with anomalously low sample count, these weak data may be
excluded for selected alarms by setting a new 'exclude_outliers'
field on the alarm threshold rule.
An outlier in this context is considered to be an aggregate with
sample count that differs from the observed mean by more than two
standard deviations.
Change-Id: I45d05db57662588df5796a8d6864ed01d1e4165f
a4c7411ac9