Docs: Clarify config.toml usage in evaluation harness

[xingyaoww]

This PR updates the documentation to clarify the usage of config.toml in the evaluation harness, fix #6813.

Changes made:

1. Updated the documentation in evaluation/README.md to clarify that only the LLM section in config.toml will be used for evaluation.
2. Explained that other configurations like save_trajectory_path are set in the get_config function of the respective run_infer.py file for each benchmark.

These changes ensure that users understand how configuration settings are applied during the evaluation process.

Update for issue #6813:
Thank you for reporting this issue. We have clarified the documentation regarding the use of config.toml in the evaluation harness. Here's a summary of the situation:

1. The evaluation harness for swe_bench doesn't read the save_trajectory_path configuration from config.toml.
2. We've updated the documentation to clarify that only the LLM section in config.toml will be used for evaluation.
3. For other configurations specific to evaluation, such as save_trajectory_path, these are set in the get_config function of the respective run_infer.py file for each benchmark.

This PR implements these documentation changes to prevent confusion in the future.

Thank you for your patience and for helping us improve OpenHands!

---

To run this PR locally, use the following command:

docker run -it --rm   -p 3000:3000   -v /var/run/docker.sock:/var/run/docker.sock   --add-host host.docker.internal:host-gateway   -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:e616cbb-nikolaik   --name openhands-app-e616cbb   docker.all-hands.dev/all-hands-ai/openhands:e616cbb