Add agent run inspection debugging tip

[luoyuctl]

Summary

• Add a Testing and Debugging tip to inspect saved AI agent runs when tasks are slow, costly, or repeatedly failing
• Mention agenttrace as one local tool for surfacing token usage, cost, latency gaps, tool failures, and retry loops
• Keep the top-level README and testing_and_debugging_guide/README.md in sync

Validation

• git diff --check

[luoyuctl]

Refreshed the debugging tip copy so both README locations mention the current local session-log coverage (Claude Code, Codex CLI, Gemini CLI, Aider, Cursor, Qwen Code, Cline, OpenCode/OpenClaw, and Kimi CLI). Verified with git diff --check upstream/main...HEAD; this repo has no CI checks configured and GitHub currently reports the PR as mergeable (CLEAN).