feat: add configurable Copilot runner

[Copilot]

Adds support for running experiments through the Copilot SDK in addition to the Copilot CLI. Experiments can now include runner choice as a dimension, making CLI vs SDK behavior comparable across the same models, scenarios, and treatments.

• Runner configuration

  • Added a Copilot runner option for experiment configs.
  • Supports selecting cli, sdk, or multiple runners to expand experiment coverage.

• Execution path

  • Preserves the existing Copilot CLI behavior as the default.
  • Adds a Copilot SDK execution path that emits compatible run output.

• Experiment dimensions

  • Includes runner selection in generated treatments/results so comparisons remain attributable.

export const experiment = defineConfig({
  name: 'Example experiment',
  description: 'Compare Copilot CLI and SDK runs',
  models: ['gpt-5.5'],
  scenarios: ['001-agent-uses-button-from-primer'],
  runners: ['cli', 'sdk'],
  treatments: [],
})