Runs all test cases against the agent. The run must be in queued status.
After execution, use the evaluate endpoint to score results.

By default (async_mode=true) this returns a V2RunActionResponse
immediately with the run in a non-terminal state — poll
GET /testing/v2/runs/{run_id} until status is executed, failed, or
cancelled. Set async_mode=false to block until execution finishes and
receive the updated V2Run.