fix: resolve dataset path config and undeclared provider dependencies

[himanshu231204]

Summary

• Dataset config bug: \config.yaml\ pointed \dataset.path\ at a non-existent \data/questions.json\ (and a bare directory \data/\ produced a misleading \InvalidDatasetError). Now points to the existing \data/sample_questions.json\ with an explicit \ormat: json.
• Clearer loader errors: \datasets/factory.py\ now detects directory paths and raises an actionable error instead of the misleading \set the format field\ message.
• JSON loader: supports the common {\items: [...]}\ wrapper so \sample_questions.json\ loads (5 items).
• Undeclared dependencies: LLM provider SDKs (\openai, \�nthropic, \groq, \google-genai, \ iktoken) were imported unconditionally but never declared. Added a \providers\ optional extra in \pyproject.toml\ (and to \�ll).
• Lazy provider imports: the four LLM provider modules now import their SDKs lazily (guarded \ ry/except), so the modules import without the optional dependency and raise a clear \ProviderError\ with install instructions at instantiation — matching the project's existing lazy-import design.
• Test fix: added the missing \import httpx\ in \ est_anthropic.py\ (was using \httpx.Response\ unimported → 3 NameError failures).

Test plan

• Full suite: 526 passed, 3 skipped (no regressions).
• Verified provider modules import cleanly without their SDK and raise \ProviderError: The '' package is required ... Install with: pip install openagent-eval[providers]\ when instantiated without it.

Notes

Only files related to this fix were committed; pre-existing unrelated modifications (e.g. README.md) and untracked artifacts (reports/, docs) were left out.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.