docs: add anchor evaluation concept and link in footer

Add evaluation framework for testing semantic anchors across LLMs:
- 4 evaluation levels (Recognition, Application, Differentiation, Consistency)
- Multiple-choice approach eliminates LLM-as-judge dependency
- Position bias mitigation via randomized option order
- 6 target models (3 commercial, 3 open-weight)
- Anchor vs. paraphrase comparison for measuring anchor value
- Concrete pilot plan and effort estimates

Integrate into website: render .adoc to HTML, add route, footer link,
and i18n translations (EN/DE).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: raifdmueller