FEAT: Add authors and group affiliations to 21 seed-dataset YAMLs

[romanlutz]

Adds authors: and groups: (affiliations) metadata to 21 seed-dataset / prompt YAMLs whose source paper is in doc/references.bib but the YAML itself was missing the affiliations. Affiliations were extracted from each paper's arXiv abstract page or PDF first page, then independently cross-checked against the papers by a second model.

Notable decisions:

• tense_converter.yaml replaces the previous authors: [AI Red Team] with the actual paper authors (Andriushchenko & Flammarion, arXiv:2407.11969).
• codechameleon_converter.yaml expands the bib shorthand Lv, Huijie, et al. to the full 9-author list with the two Fudan University affiliations.
• many_shot_template.yaml uses the full 34-contributor list and 6 institutions from the Anthropic technical report (Anthropic, University of Toronto, Vector Institute, Constellation, Stanford, Harvard) rather than collapsing to [Anthropic].
• dt_stereotypes_*.yaml (3 files) lists all 19 DecodingTrust authors and 6 institutions.
• Affiliations use institution-name form only (no , Country address suffixes); UC campuses use the short form (UC Berkeley, UC Davis) consistently across all files.

All 21 YAMLs parse as SeedPrompt. Related unit tests pass (75 tests across test_jailbreak_text, test_flip_attack, test_many_shot_template, test_many_shot_jailbreak, test_persuasion_converter, test_code_chameleon_converter).