Reduction over different columns of a single DataFrame can be merged

When calculating series aggregations like `execute(df.a.sum(),df.b.mean())`, aggregations over different columns can be merged as `df.agg({'a': 'sum', 'b': 'mean'})`. An optimizer can be added to reduce num of subtasks.

Start by tracing how execute handles separate series aggregations such as df.a.sum() and df.b.mean(), then compare that path with df.agg({'a': 'sum', 'b': 'mean'}). Identify where reduction subtasks are planned and determine whether compatible aggregations can be combined. Done means equivalent results with fewer subtasks.

Written by the indexing model from the issue text.