Generation jobs

A generation job validates one authored graph template, schedules deterministic run/sample/shard seeds, executes locally or through Dataproc, and publishes through the configured rfgen.storage.Storage implementation. The built-in templates are available through rfgen init.

Dataproc workers reconstruct both the executor and storage implementation from the submitted configuration. A job publishes the root only after every shard has succeeded; partial runs remain unpublished.