Remove duplicate research source supplement

[erinepshovel-code]

Purpose

Repairs the research-ledger uniqueness regression currently failing main and the synthetic merge gate for PR #17.

src/_data/research/source_supplements.yml repeats four source IDs that already exist in the primary reviewed ledger:

• un-crpd-2014-legal-capacity
• stephan-chenoweth-2008-civil-resistance
• rosch-1975-semantic-categories
• solove-2013-consent-dilemma

The duplicate records cause the executable source-identity invariant to fail (21 !== 25). All claims already resolve through the primary sources.yml, so no research evidence is removed.

What changed

• removes the duplicate supplemental source file
• removes the now-unused optional supplement-loading path from the Eleventy research adapter
• restores the test to the single authoritative source ledger
• preserves source-ID uniqueness rather than weakening or bypassing the guardrail

Evidence boundary

This is a data-identity repair only. It does not alter any research claim, stance, limitation, Article Lab content, canonical text, or external source.

Validation

The branch differs from current main in exactly three research-ledger files. The normal pull-request workflow will run the complete build, research, generated-site, browser, and accessibility gates.

hmmm

The four repeated records contained slightly different metadata wording, but not four additional sources. Retaining both versions would manufacture corpus size and make one source appear twice; the primary reviewed records remain authoritative.