Better support fo multi-col predicate evaluation

With #94 we will be able to run query like this:

select
    o_orderpriority,
    count(*) as order_count
from
    orders
where
        o_orderdate >= '1993-07-01'
  and o_orderdate < date '1993-07-01' + interval '3' month
  and exists (
        select
            *
        from
            lineitem
        where
                l_orderkey = o_orderkey
          and l_commitdate < l_receiptdate
    )
group by
    o_orderpriority
order by
    o_orderpriority;

Where the predicates may use multiple columns: l_commitdate < l_receiptdate.

The current approach is to convert the columns to arrow and then perform the arrow predicate eval.

This means that we don't get the benefit of evaluate on encoded data. I think we can do better here.

Start by tracing the current path that converts columns to Arrow before predicate evaluation, using the multi-column predicate example in the issue as the reference case. Compare that path with evaluation on encoded data; done means predicates involving multiple columns, such as l_commitdate < l_receiptdate, are evaluated correctly without losing the encoded-data benefit.