Build a small `bg_job_runner(tasks)` that ties 4 module-3 ideas together: Each `task` is a dict with `{"id": str, "work": int, "fail": bool}`. The runner must: 1. **Iterate deterministically** — process in list order so tests pass. 2. **Retry up to 2 times** on tasks with `fail=True`. On the 3rd attempt the task succeeds (simulating a flaky dep that eventually responds). 3. **Track each task's final state** — `{"id": id, "attempts": n, "result": work*10}`. 4. **Return** the list of results in the original order. (We avoid real `async` here so the recap stays Pyodide-friendly; the retry semantics are the actual idea you carry into module 4.)