After #144: point licence-review links at #35, record the pandas_panel translation readers, fix two stale consumer notes

[mmcky]

The 2026-09-29 close-out review of the datasets migration found four small corrections. Each one changes a manifest note or consumers list, so CATALOG.md has to be regenerated with scripts/build_catalog.py. That regeneration would conflict with open #144, so these wait until it merges and then land as one PR.

• Licence-review links point at a private issue. The catalog legend (scripts/build_catalog.py), lectures/countries.csv.yml (two places) and AGENTS.md's licensing paragraph all send readers to QuantEcon/workspace-lectures#20. That issue is private, so the public gets a 404. It was folded into Licensing: record-and-track policy, and the per-dataset inventory for migration #35 on 2026-09-29 and closed as a duplicate of it, so point all four at Licensing: record-and-track policy, and the per-dataset inventory for migration #35.
• The pandas_panel files have three unrecorded readers. As measured on 2026-09-29, lecture-python-programming.zh-cn, .fr and .fa read countries.csv, employ.csv and realwage.csv from this repo. The three manifests list only lecture-python-programming, lecture-python.myst and lecture-python.zh-cn. test_pwt.csv.yml already lists the same translations, so add them to all three consumers lists.
• us_adult_heights.csv.yml has a stale consumer note. It says lecture-intro.zh-cn's prob_dist.md reads a local _static/lecture_specific/prob_dist/ copy. In fact it reads this repo's copy (github.com/QuantEcon/data-lectures/raw/main/lectures/us_adult_heights.csv, lines 75–76 of that file), and the local folder no longer exists. Drop the note.
• dataBHS.csv.yml has a stale consumer note. It says lecture-tools-techniques reads its own dataBHS.mat until five_preferences: downloadable notebook cannot run — dataBHS.mat is not served by the site lecture-tools-techniques#11 lands. That change landed through five_preferences: read dataBHS from data-lectures, guard usetex for Colab lecture-tools-techniques#12 on 2026-08-19, and five_preferences.md now reads dataBHS.csv from here, so rewrite it as a plain reader entry. Whether lecture-tools-techniques should delete its now-unread lectures/dataBHS.mat is that repo's call.

Before merging: python scripts/validate_datasets.py passes, and after the commit, running python scripts/build_catalog.py leaves no diff.

Found during the migration close-out, tracked in QuantEcon/workspace-lectures#14 (private). The related docs tidy is #145.

[mmcky]

Two more small corrections for the same PR, found on 2026-09-30 while reviewing #152 and the canary proof (#147). Neither needs a CATALOG.md regeneration.

• manifest-schema.yml line 113, in the license.note comment: "In use by 16." is stale, since 27 manifests set license.note on main today. It is a count that drifts with every manifest, so delete the sentence rather than update the number. Lines 79, 84 and 92 carry the same kind of count; I did not check those.
• .github/workflows/refresh-snapshots.yml line 162, the notify job: the issue body opens "The weekly sources-alive canary failed", which is wrong for a dispatched run (seen on refresh-snapshots: a dynamic snapshot's builder is failing #150). "The sources-alive canary failed" fits both triggers.