Auto-annotate images or blur faces

For auto-annotation, saves a dataset version (cloud storage datasets have none), then queues a run that labels the dataset's unlabeled images with the given model, or every image when includeAnnotated is set. Set operation: blur to blur faces using the platform detector, optionally limited to imageId. Blurring requires hosted storage and creates no version. confidence defaults to 0.25; boxScale defaults to 1 and scales face boxes around their centers. Set preview: true to prepare up to six full-resolution images and their thumbnails without changing the dataset. Pass the returned jobId as previewJobId with the same settings to apply those exact assets; only remaining dataset images require processing. Existing labels are never changed, and the run is billed for the images it actually processes.