fix: improve API error handling with detailed diagnostics

[roomote]

Summary

This PR attempts to address Issue #9888 by improving error handling and diagnostics for API requests. The changes provide more detailed error information to help debug intermittent "API request failed" errors.

Changes

Enhanced Error Messages

• Added comprehensive error diagnostics including provider, model, endpoint, and timestamp
• Improved error messages for both first-chunk failures and mid-stream failures
• Added request metadata (attempt number, token counts, error codes) to error messages

Better Retry Logic

• Enhanced retry logic with exponential backoff
• Added detailed logging for debugging intermittent failures
• Improved handling of empty assistant responses with auto-retry capability

Request Diagnostics

• Added diagnostic information for streaming failures including stream position and partial content status
• Enhanced logging with structured error data for better debugging
• Added token usage tracking in error diagnostics

Related Issue

Fixes #9888

Testing

• All existing tests pass (342 test files, 4527 tests passed)
• No breaking changes to existing functionality
• Error handling improvements are backward compatible

Feedback Welcome

This PR attempts to address the intermittent API errors reported in the issue. Feedback and guidance are welcome!

[roomote]

Rooviewer   See task on Roo Cloud

Review complete. No issues found.

This PR adds comprehensive error diagnostics to help debug intermittent API failures. The changes include detailed error context (provider, model, timestamps, token usage, error codes) for streaming failures, empty responses, and first-chunk failures. All 4527 tests pass and the implementation follows existing error handling patterns.

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}