Update platform to use upcoming asynchronous ML backend API

[mihow]

See Kavi's documentation

[mihow]

I have made good progress on this task. The platform has been reworked to accept callbacks and initiate async jobs. With our first image successfully processed by flatbug and the order classifier:

[mihow]

Claude says: Closing as completed. The asynchronous ML backend (PSv2) is now the platform default as of #1353 (merged, deployed 2026-06-27): async_pipeline_workers defaults to True and all existing projects were migrated over. The async dispatch path (JobDispatchMode.ASYNC_API — NATS queue, worker pull, result ingestion, cancellation, monitoring) shipped across #987 / #1046 / #1146 / #1162 / #1227 / #1244 and is now the common path. Remaining PSv2 work is tracked as non-blocking follow-ups under the umbrella #515.