Copilot and Qwen coding agent

[Cedric / ViaDézo1er (viadezo1er)]

Add copilot and qwen to the agents supported by bt setup
There are commits related to Qwen as this branch was created from cedric/add-qwen-coding-agent before it was merged into main.

Small DRY refactor.

Bug fixes:

• gemini doesn't trust the folder by default when launched in background mode, which is a pitfall if used in CI (WIP)
• Cursor and gemini need --output-format stream-json to output their thinking in real time in background mode, which requires some custom code.

[github-actions]

Latest downloadable build artifacts for this PR commit 60be5e9fe078:

• Workflow run: https://github.com/braintrustdata/bt/actions/runs/25180766898
• Download all artifacts (GitHub CLI): gh run download 25180766898 --repo braintrustdata/bt
• Installers are published from main automatically. To publish one for a PR branch, run release-canary manually via workflow_dispatch.

Available artifact names

• ``artifacts-build-global
• ``artifacts-build-local-x86_64-apple-darwin
• ``artifacts-build-local-x86_64-pc-windows-msvc
• ``artifacts-build-local-aarch64-apple-darwin
• ``artifacts-build-local-x86_64-unknown-linux-musl
• ``artifacts-build-local-x86_64-unknown-linux-gnu
• ``artifacts-build-local-aarch64-unknown-linux-gnu
• ``artifacts-plan-dist-manifest
• ``cargo-dist-cache

[Cedric / ViaDézo1er (viadezo1er)]

Example config to test qwen
~/.qwen/settings.json:

{
  "env": {
    "OPENAI_API_KEY": "a real openai api key"
  },
  "modelProviders": {
    "openai": [
      {
        "id": "gpt-5-nano",
        "name": "GPT-5-nano",
        "envKey": "OPENAI_API_KEY",
        "baseUrl": "https://api.openai.com/v1",
        "generationConfig": {
          "timeout": 60000,
          "maxRetries": 3,
          "enableCacheControl": true,
          "contextWindowSize": 128000,
          "modalities": {
            "image": true
          },
          "customHeaders": {
            "X-Client-Request-ID": "req-123"
          },
          "extra_body": {
            "service_tier": "flex"
          },
          "samplingParams": {}
        }
      }
    ]
  },
  "security": {
    "auth": {
      "selectedType": "openai"
    }
  },
  "model": {
    "name": "gpt-5-nano"
  },
  "$version": 3,
  "mcpServers": {
    "braintrust": {
      "httpUrl": "https://staging-api.braintrust.dev/mcp",
      "headers": {
        "Authorization": "Bearer sk-****a real braintrust api key"
      }
    }
  }
}

[Abhijeet Prasad (AbhiPrasad)]

Why did we introduce this? Now the sync and async tests can be run in parallel, and access env, which is bad. We should have a single shared mutex.