[Features] More Aggregations

[jaychia]

Summary

Aggregations are used in a few contexts:

1. Global aggregation (performed on the entire column of a DataFrame, or semantically equivalently - grouped by an empty set of keys [])
2. Grouped aggregations (performed on columns grouped by certain keys)
3. Windows (rolling windows of computation)

Daft currently supports a few aggregations: sum, count, mean, list, concat, min, max.

Some of the aggregations we would like to support in the future include the following:

• std
• set - like list, but only keep distinct elements
• count_distinct - like count, but only count distinct elements
• quantiles