I need to add a new column to a large production table and backfill values for existing rows. Write me a safe backfill plan and the script to execute it. Requirements for the backfill: 1. Batch: process rows in small batches (a few hundred to a few thousand at a time) keyed by primary key ranges, not OFFSET, so each batch is fast and the table is never locked for long. 2. Resumable: track progress (last processed ID) so the script can be killed and restarted without redoing work or double-writing rows. Make each batch idempotent. 3. Throttled: add a small delay or row-per-second limit between batches so the backfill never competes with production traffic. It should be safe to run during the day. 4. Observable: log progress per batch (rows done, estimated remaining) so I can watch it. 5. Compatible: the deploy order matters. The new column must be nullable (or have a default) so old and new code both work while the backfill runs. Application code should read the new column with a fallback to the old value until the backfill completes, then we tighten it in a later migration. Do not write a single UPDATE that touches the whole table at once; that is the outage we are avoiding. Do not add a NOT NULL constraint until after the backfill finishes and I confirm it. Give me: the migration for the column, the backfill script, a dry-run mode that reports how many rows need filling without writing, and the follow-up migration to enforce constraints once the data is in.