perf: reduce background CPU churn and battery usage.

[StarsExpress]

Refs issue #438 to decrease busy-wait polling in code_puppy/messaging/bus.py: get_message(), get_command().

---

PR #859's choice

Consider a real handoff: keep an asyncio.Queue bridged via call_soon_threadsafe (as message_queue.py already does), or at minimum back off the sleep when idle.

PR #859 implements the latter option — back off sleep interval when idle — as an alleviation for background CPU churn and battery usage.

Full fix to bridge code_puppy/tools/command_runner.py's background threading.Thread — used for streaming subprocess output — into the event loop via loop.call_soon_threadsafe is a larger and thus separate follow-up, since a straight swap to asyncio.Queue would break these cross-thread writes.

---

Alleviation implementations

code_puppy/messaging/bus.py: MessageBus now has a shared static method for get_message() and get_command():

T = TypeVar("T")

@staticmethod
async def _get_nowait_with_backoff(q: "queue.Queue[T]") -> T:
    """Poll a thread-safe queue with exponential backoff while idle.

    Wraps a sync queue in an asyncio-friendly way. The poll interval
    starts at 0.01s and doubles on each empty attempt up to a cap of
    0.1s, resetting to 0.01s on the next call (i.e. once a value is returned).
    """
    delay = 0.01

    while True:
        try:
            return q.get_nowait()

        except queue.Empty:
            await asyncio.sleep(delay)
            delay = min(delay * 2, 0.1)

async def get_message(self) -> AnyMessage:
    """Get the next outgoing message (async).

    Called by the renderer to consume messages.
    Blocks until a message is available.

    Returns:
        The next message to display.
    """
    return await self._get_nowait_with_backoff(self._outgoing)

async def get_command(self) -> AnyCommand:
    """Get the next incoming command (async).

    Called by the agent to consume commands (e.g., CancelAgentCommand).
    Blocks until a command is available.

    Returns:
        The next command to process.
    """
    return await self._get_nowait_with_backoff(self._incoming)

---

Idle CPU wakeups reduction

Before PR #859	Inside PR #859	Reduction %
454	52	88.5%

---

Tradeoffs

After empirically measuring for a message arriving after 0.4s of idle, additional latency introduced by this backoff is ~57.8ms.

Theoretical worst case is bounded by 0.1s backoff cap itself, since delay resets to 0.01s immediately after each successful read.

---

Additional adjustments

Inside code_puppy/messaging/bus.py, I added some blank lines to visually separate code chunks for better readability.