Read token usage from responses API

[anelephantbowl]

Note

Make LiteLLM token usage parsing support both Responses and Chat Completions APIs, update tests, and bump version.

• Metrics (LiteLLM):
  • Adjust '_get_token_usage' in metrics/eval_clients/_litellm.py to read token usage from both Responses API (input_tokens/output_tokens) and Chat Completions (prompt_tokens/completion_tokens).
• Tests:
  • Update tests/metrics/eval_clients/test_litellm.py to mock ResponseAPIUsage and validate summed token usage for Responses API scenarios.
• Version:
  • Bump version to 0.10.0.dev13 in pyproject.toml and src/langcheck/__init__.py.

^{Written by Cursor Bugbot for commit 12e6b38. This will update automatically on new commits. Configure here.}

[taniokay]

Thanks! Confirmed that it correctly capture the tokens for reasoning too!