geojson-flatten

I run Live Tennis API. Collection flattening repeatedly concatenates each result onto the accumulated array, copying earlier results again for every member. This becomes expensive for large FeatureCollections and GeometryCollections. Use for both collection paths and add a patch changeset. Existing output order and input mutation behavior are preserved. Verified on Node 22.23.2 with the locked pnpm 12.4.1 dependencies: : Biome and all 7 tests pass. 201 output/input-mutation comparisons against pass, including nested collections, null geometries/properties, feature IDs, empty parts, and a 200,000-point part. Output digests match for 5,000, 10,000, 20,000 and 40,000-member collections. Local timings for 40,000 point members (median of three runs, cloning outside the timed section): Timing reproduction Run this as an ES module from the repository on main and on this branch:

Doesn't require a changeset, this is just internal.

Fixes #103