Combine several standardizations into a single new standardization on a target document.

This exists for split documents. A long file is split into children, each child is standardized on its own so they process in parallel, and this call reassembles them into one result on the parent document - so anything downstream sees one extraction per source file rather than one per child.

List fields are concatenated in the order the standardization ids are supplied, so pass them in page order. Non-list fields take the first non-null value. Where the children were standardized against different schemas, the merged standardization records the most common one.

The result is a new standardization, retrievable and transformable like any other; the originals are left untouched.

No extraction runs, so the cost is a flat 1 credit per call regardless of how many standardizations are merged or how many rows they hold.