Dependent metric equality is fragile on new ProtocolMetricImpl

[michael-mclawhorn]

1st- The static reference to this.dependentMetrics can be overshadowed in suprising ways in subclasses through the static inclusion.
2nd - Equality on dependent metrics is so strict as to be problematic for testing and any kind of proxying. It should be sufficient to note that the dependent metric names are identical strongly indicating the same meaning with less test burden.