Metrics: Investigate using metrics key with bucket index for histograms

[tsloughter]

Instead of bucket_counts being a counters array which will be able to get out of sync with the min/max values stored separate in the ets element we should see about changing the ets table for histograms to use a key of {Key, BucketIdx}.

Then either a separate element with the min/max is kept or could even keep min/max per-bucket and aggregate them during collection I think. That way only a single element has to be updated when a bucket is incremented.

I don't think this has to be implemented before we cut GA of metrics, but it should be investigated to make sure it is possible with the current design and won't change too much post-GA to implement it.