Back-fill articles.semantic_embedding across full DEV corpus

Is your feature request related to a problem? Please describe.

As a DEV admin who utilizes articles.semantic_embedding in reporting contexts, I'm currently limited to the set of articles with completed embeddings (not the full published archive). Coverage today is ~82k of ~2.3M published articles (~3.6%), and it's recency-skewed — embeddings are generated going forward on publish, so most of the historical corpus was never embedded.

Describe the solution you'd like

Back-fill articles.semantic_embedding across the full published archive — a one-time, chunked, resumable sweep of the ~2.25M unembedded published articles, using the same embedding model/dimension as the existing column.