Improve Tuple sketches docs

From @gianm :

Think of a tuple sketch as a theta sketch plus an additional vector of numeric values associated with each item being added to the sketch. The items added to the sketch are counted, intersected, etc just like a theta sketch. Where it goes beyond the theta sketch is that you can get the mean, variance, sum of the fields in the vectors.

For example: you might be adding user IDs to the theta sketch, and you might have a lifetime value for each user, and you might want to compute the average lifetime value for users matching a certain set of parameters. You can do this with tuple sketches. The added vector would be a 1-vector (lifetime value).

> Is the use case comparable to quantile sketches?
> Not really. It's more of an extension of theta sketches.

No file or test is named in the issue. Start by locating Druid's tuple-sketch documentation; done means clearly explaining the tuple-sketch model, the lifetime-value example, and how it differs from theta and quantile sketches.

Written by the indexing model from the issue text.