A machine learning engineer needs to create a pipeline that runs a custom container component on Vertex AI. The container expects a Cloud Storage path as input and outputs a model artifact. Which component type should they define using the Kubeflow Pipelines SDK v2?