Improve Test Coverage for Tool Use Response Handling

[jchris]

The current tool use response handling code has either insufficient test coverage or contains potentially dead code paths. We need to ensure all branches are properly tested, especially for different API response formats when handling tool use responses.

Issues

1. Low test coverage for different tool use response formats, particularly for Claude responses
2. Multiple code paths might not be exercised by current tests
3. Unclear if all these conditionals are necessary or if some are obsolete

Action Items

• Verify accuracy of test fixtures for Claude wire that cover tool use responses
• Add test cases for OpenAI/other providers with tool use responses
• Test all branches in the tool use response handling code
• Verify if all conditionals are necessary or if some can be simplified/removed
• Document expected response formats from each provider when using tool mode

Additional Context

The coverage report shows this section isn't fully tested. We need to determine if this is due to lack of test fixtures or if some of these conditions are no longer relevant to current API response formats.