Based on @andrewkaufman's comments on https://github.com/newton-physics/newton/pull/996, we should rethink the `SchemaResolver` implementation and make sure the schemas encoded in them are correct.