feat(news): licence third-party bias ratings from AllSides or Media Bias/Fact Check

## What Ground News actually does
Kal described, without naming it, an app showing every country's and language's
framing of a story with a closeness-to-truth ranking. That is **Ground News** —
and usefully for a Toronto customer, it is a **Canadian company based in
Kitchener, Ontario**, founded 2018 by an ex-NASA engineer.

The key finding: **Ground News does not rate sources itself.** It aggregates
**three third-party raters — AllSides, Ad Fontes Media, and Media Bias/Fact
Check.** Bias weighs "wording, story choices and political affiliation";
factuality weighs "the credibility of sources used, the speed at which
corrections are made, and whether the language retains context."

Its real moat is **2,276 outlets hand-coded for ownership and funding** — that
is proprietary and expensive. The bias ratings are licensed and therefore
replicable.

Scale: ~60,000 articles/day from 50,000+ sources. Pricing: Vantage at $8.33/mo
billed annually, with a free tier. **No API or data-licensing offering is
advertised.**

## Recommendation
**License a named third party rather than inventing a scale.** Inventing one
invites an argument we cannot win and contradicts our transparent-sourcing
position; citing AllSides or MBFC is consistent with it — we would be showing
our work, exactly as we do with source provenance.

This is also days of work, not months: license or reproduce a source-rating
table and join it to article-source metadata we already carry. The presentation
layer (per-source colour chip, left/centre/right coverage split, blindspot
framing) is a UI problem, not a data problem.

## Acceptance
- [ ] Licensing terms established with at least one rater
- [ ] Ratings joined to source metadata
- [ ] Displayed with clear attribution to the rater
- [ ] Feeds the credibility score (separate issue)

Related: #1994 (user-defined Telegram channel watchlist) is adjacent — both
touch how per-source trust is assigned and surfaced.

Effort: L (~7d) including licensing negotiation.

No files, tests, or entry points are named. Start by establishing licensing terms with AllSides, Ad Fontes Media, or Media Bias/Fact Check, then inspect the existing article-source metadata and credibility-score flow. Done means ratings are joined to source metadata, shown with clear attribution, and passed into the separate credibility score.

Written by the indexing model from the issue text.