Add AgentCompass to evaluation and testing resources

[natnew]

Motivation

• Add AgentCompass to the Evaluation and Testing section because its modular Benchmark–Harness–Environment design, asynchronous execution, and trajectory-level failure analysis directly address repeatable agent evaluation, failure diagnosis, and observability needs in AgentOps.

Description

• Add a single README entry under Evaluation and Testing: - [AgentCompass](https://arxiv.org/abs/2607.13705) - Evaluation infrastructure for agent benchmarks, harnesses, environments, asynchronous execution, and trajectory-level failure analysis. Codex was used to review the repository contribution guidance, verify the canonical resource and link, craft the README edit, and prepare this PR description by generating the concise entry and validation steps.

Testing

• Verified the repository guidance and relevant files were read using file inspections of README.md, CONTRIBUTING.md, and SECURITY.md to ensure the change follows local contribution rules and formatting conventions.
• Performed an automated duplicate search with the pattern AgentCompass|2607\.13705|open-compass and found no existing matching entries.
• Confirmed the arXiv landing page https://arxiv.org/abs/2607.13705 resolves; no automated test suite or linter was run because this is a documentation-only change.

---

Codex Task