Improve Gemini API retry logging

[Shashank23123458867504]

Pull Request

Linked issue

Closes #699

What & why

This PR improves the retry helper used by the Google embedding provider.

Changes

• Refactored retryable status detection using any() over a centralized tuple of retryable status codes.
• Improved retry logging by including the exception type.
• Added debug logging for non-retryable exceptions.
• Preserved the existing retry behavior and exponential backoff logic.

These changes improve the readability and maintainability of _call_with_retry() without changing its intended behavior.

How it was tested

The local development environment has not yet been fully configured, so I was not able to run the project's test suite, linting, or type checking. The change is localized to _call_with_retry() and preserves the existing retry semantics.

Checklist

• Tests added for new functionality
• All tests pass: uv run pytest tests/ --tb=short -q
• Linting passes: uv run ruff check code_review_graph/
• Type checking passes: uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
• Lines are at most 100 characters
• Docs updated where behaviour changed (README, docs/, docstrings)