Group by metric when aggregating

[jc-harrison]

We could consider adding a 'groupby' parameter to joined spatial aggregate queries (and non-spatial aggregates - c.f. #2022), to specify a categorical metric to group by before aggregating. Query results would then have columns [<location_id_col(s)>, group, value] instead of just [<location_id_col(s)>, value] (similar to the current method="distr" option, but values would be aggregates of a second metric, instead of proportions).

We could also consider exposing a way of binning continuous metrics, so they could be used for grouping. If so, bins should be consistent with those obtained from a histogram aggregeation.