Support "post-aggregation" expressions (or a multi-weight ratio-delta aggregate) for row-pivoted views

Description of Problem: Expression columns are evaluated at leaf granularity and then aggregated with the column's chosen aggregate. This works well for a single ratio via the built-in weighted mean aggregate (e.g: #2324).

aggregates: { "conversion": ["weighted mean", "traffic"] } → correctly yields sum(customers)/sum(traffic) at every group_by level.

However, there's no way to correctly aggregate a derived metric built from two independently-weighted ratios with different weight columns, e.g.:

conversion = customers / traffic conversion_py = customers_py / traffic_py conversion_delta = (conversion - conversion_py) * 100

No single built-in aggregate can express "weighted mean of x with weight w1 minus weighted mean of y with weight w2" — the result is not a weighted mean of (x - y) under any single weight vector when w1 and w2 differ per row.

Potential Solutions:

1. Allow expressions to be evaluated after aggregation, operating on already-aggregated column values within a pivoted view (i.e. a "post-aggregation expression" concept), or
2. A new built-in aggregate that accepts two (numerator, denominator) pairs and computes the difference of the two resulting ratios at each pivot level, or
3. Support composing a View's aggregated output directly as input to a new set of expressions without round-tripping through a second Table.

Happy to provide a minimal repro dataset/config if useful.