`status` and `iterate` re-run `babs status` per active cell with no `--no-refresh` and no progress

`mechababs status` shells `babs status --json` once per active cell (`status.py`, via `babs_status.read_status`), and babs recomputes `job_status.csv` from the scheduler on every call. At a superstudy that is one subprocess per active cell, sequential and silent: on the c5 wave (56 cells) `status` is slow and opaque. `iterate` routes off the same call, so it pays the same cost every tick.

Wanted, matching what `jobs` already has:

- `--no-refresh`: read each cell's `job_status.csv` as it stands; the header says the table was not refreshed.
- a per-cell progress line on stderr while refreshing.

Related, from the same session: `active` covers both scaffolded-but-unsubmitted and running. The `jobs` column disambiguates, but `state` should say it.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Start in status.py at babs_status.read_status and compare the existing jobs handling for refresh suppression and output. Trace how status and iterate invoke the per-cell refresh, then verify --no-refresh reads the existing job_status.csv, refreshes show per-cell stderr progress, and state distinguishes scaffolded-but-unsubmitted from running.

Written by the indexing model from the issue text.