Add Vectara Hallucination Leaderboard adapter

[mohammadrezakarami]

Add Vectara Hallucination Leaderboard adapter

Summary

Adds a one-off utility adapter for parsing selected Vectara Hallucination
Leaderboard result files from the public vectara/results Hugging Face dataset
into the Every Eval Ever schema.

This is the companion code for the datastore submission:

https://huggingface.co/datasets/evaleval/EEE_datastore/discussions/144

Details

• Adds utils/vectara_hallucination_leaderboard/adapter.py
• Pins source result fetching to commit
  7c104699e98ade53dd719f79ae9f7eb281c8107d
• Keeps the private/proprietary underlying evaluation dataset separate from the
  public structured result source (vectara/results)
• Records inference platforms for the selected models:
  vertex_ai, azure, and dashscope
• Includes offline source row snapshots for deterministic smoke tests
• Adds tests for schema-valid export, scores, platforms, and source-data
  semantics

Validation

PYTHONPATH=/private/tmp/eee_code_pr_deps python3 -m pytest tests/test_vectara_hallucination_adapter.py
# 4 passed

PYTHONPATH=/private/tmp/eee_code_pr_deps python3 -m every_eval_ever validate /private/tmp/eee-vectara-code-pr-smoke
# All 3 file(s) passed validation