Computes `lengthkm` on flowpaths and `areasqkm` on divides, then joins each divide's incremental `areasqkm` onto its flowpath. Any `areasqkm` already on `flowpaths` is dropped first, so the returned value always comes from `divides`. The geometry column of both layers is renamed to `"geometry"`.
Details
The divide-to-flowpath join prefers an explicit `flowpath_id` column on `divides`, which the current schema requires because `divide_id` and `flowpath_id` use different namespaces (`cat-*` vs `fp-*`). When `divides` carries no `flowpath_id`, the join falls back to the legacy 1:1 convention that `divide_id == flowpath_id`. Picking the wrong key here matches nothing and silently zeroes `areasqkm`, along with any total area accumulated from it, so a fabric whose divides lack `flowpath_id` under the current schema should be repaired rather than passed through this fallback.
