Share state between Modal containers using modal.Dict and modal.Queue

Documented steps

1. Create named, persistent objects: `d = modal.Dict.from_name("job-state", create_if_missing=True)` and `q = modal.Queue.from_name("work", create_if_missing=True)`.
2. For scratch state scoped to one script run, use the context managers instead: `with modal.Dict.ephemeral() as d, modal.Queue.ephemeral() as q:` — both are destroyed on exit.
3. Dict operations: `d.put(key, value, skip_if_exists=False)`, `d.get(key, default)`, `d.pop(key, default)`, `d.contains(key)`, `d.len()`, `d.update(mapping)`, `d.clear()`, and the iterators `d.keys()`, `d.values()`, `d.items()`.
4. Queue producer side: `q.put(item)` or `q.put_many([...])`. Consumer side: `q.get()` or `q.get_many(n_values, timeout=5)`.
5. Wrap blocking gets in a try/except: on timeout the SDK raises `queue.Empty` rather than returning None or an empty list.
6. Pass the Dict/Queue handle into remote work with `some_fn.spawn(q, d, batch)` — the handles are serializable and resolve inside the container.
7. Use partitions to keep independent streams in one Queue: `q.put(v, partition="tenant-a", partition_ttl=86400)` and `q.get(partition="tenant-a")`. Length and clear are per-partition unless you pass total=True / all=True.
8. Reference: https://modal.com/docs/guide/dicts-and-queues