Provide a wider range of aggregate functions

Currently, we support a limited range of aggregate functions aggregation functions: avg, sum, min, max, count.

This range should be extended to support a wider range of statistical aggregates, adding for example:

* standard deviation
* rate of change
* moving-window averages
* percentile of various ranks, e.g. p99 for API responsiveness

It would also be useful to support a capabilities-style operation to allow caller discover the available aggregation functions.
(Especially since aggregates might be specific to individual storage-drivers, e.g. where some aggregate can be very efficiently calculated via mongodb MapReduce, but prohibitively expensive for say sqlalchemy).