Provide a better way to visualize the shape of the metrics data

[evanh]

Currently it is difficult to make any decisions on how to change the data store since we don't fully understand how the metrics data is actually laid out.

For example, Sentry is not a typical example of a customer. How many customers are actually exceeding the bucket size that generates a cost savings (e.g. for quantiles, if a row has less than 8192 values in it, then we might as well store raw values).

It's also difficult to understand questions like "how many more rows does the 10 second granularity store vs. the 60 second granularity?"