Should sample# map to the librato multi-sample measurements?

[mikehale]

Today sample# expects a simple value. However it seems to make sense that if sample is a summary statistic that it would have a count, sum, min, max, and min as the multi-sample gauge measurements do. They could probably also submit a perc95, perc99. Without this feature the sample generator is forced to pick one of these values.

I don't actually have a concrete use case for this at the moment, but it just occured to me as something that is or may be missing in l2met. My desire is for correctness and completeness. Let me know if this is off base.