docs: state the code/data license split explicitly

The License section read "MIT. Ground-truth annotations derive from the
Vesuvius Challenge open data release" — incomplete on both counts. It did
not mention that data/pherc1667_merged_columns/ is CC BY-NC 4.0, and it
attributed all ground truth to the open data release when the 1667 column
coordinates and transcription come from the Angelotti et al. preprint.

meta.json already stated the split correctly ("ScrollGT code is MIT; this
data directory inherits CC BY-NC"), so this aligns the front page with the
per-target metadata rather than changing any licensing fact. Matters
because the repo is a prize submission artifact and the prize terms
require a permissive license for the method: the code is MIT, and the one
non-commercial directory is third-party scholarly work that cannot be
relicensed, so it needs to be visible up front.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: jonmarrs

README.md

@@ -147,5 +147,13 @@ surface volumes: `s3://vesuvius-challenge-open-data/` (anonymous).
 
 ## License
 
-MIT. Ground-truth annotations derive from the Vesuvius Challenge open data release;
-see the challenge's data terms.
+**Code: MIT** (see `LICENSE`).
+
+**Data: mixed, per target.** Check each target's `meta.json` before reuse.
+
+- The three pixel-level Scroll-1 targets (`scroll1_*`) register 2023 Grand-Prize-era human ink
+  annotations from the Vesuvius Challenge open data release; see the challenge's data terms.
+- `data/pherc1667_merged_columns/` is **CC BY-NC 4.0**: its column coordinates and
+  transcription facts derive from Angelotti et al., *Complete virtual unwrapping and reading of
+  a rolled Herculaneum papyrus* (<https://scrollprize.org/pdf/main.pdf>). Attribution required,
+  **non-commercial use only**. This directory does not inherit the MIT grant.