Fix ma_device_stop for asynchronous backends

[MrGuim]

In the audio thread of asynchronous backends, the following condition keeps the for loop running while it waits for a new operation in the queue.

if (ma_context_is_backend_asynchronous(ma_device_get_context(pDevice))) {
    continue;   /* <-- This just makes the audio thread wait for a new operation to arrive, like an uninit or stop. */
}

You should then stop the backend when a MA_DEVICE_OP_STOP operation is queued rather than only triggering a signal.

for (;;) {
    // You will be waiting again for a new operation to be queued.
    result = ma_device_op_queue_next(&pDevice->opQueue, MA_BLOCKING_MODE_BLOCKING, &pOp); 
    if (result != MA_SUCCESS) {
        break;
    }

    if (pOp->type == MA_DEVICE_OP_UNINIT) {
        ...
    } else if (pOp->type == MA_DEVICE_OP_START) {
        ...
    } else if (pOp->type == MA_DEVICE_OP_STOP) {
        // <-- You should handle the stop operation here and not only send a signal.
        ma_device_op_do_stop(pDevice, pOp->pCompletionEvent);
    }
}

I might be missing something about how the backend interaction works, but I don't see where the backend is being stopped elsewhere

[mackron]

Thanks. Just acknowledging that I've seen this. Will report back when I get a chance to properly review this.