Add Expr.list.agg for aggregating list contents

[orlp]

This expression would take one argument, a list expression (a concept we don't have yet, see #9189 (comment), essentially just an expression in which you can only use pl.element()), which must return a scalar.

Then this would evaluate the expression for each list, returning the scalar. For example, to get the difference between the minimum and maximum in each list:

df.select(pl.col.l.list.agg(pl.element().max() - pl.element.min()))

Semantically this would be identical to the current-day:

df.select(pl.col.l.list.eval(pl.element().max() - pl.element.min()).list.first())

but it's cleaner to just write agg and allows us to ensure that you do indeed return a scalar for each list.