Allow the use of cpu when having GPUs

Describe the solution you'd like
In some cases with large models as total segmentator, or when we have large full body CT running inference can result in OOM errors using gpus. It would be nice to have the ability to pick cpu from device list

we can change the lines of https://github.com/Project-MONAI/MONAILabel/blob/main/monailabel/utils/others/generic.py#L192-L200

to

def device_list():
    if torch.cuda.is_available() 
        devices = [] 
    else:
        return ["cpu"]
    if torch.cuda.device_count() == 1:
        devices.append(torch.cuda.get_device_name(0))
    else:
        for i in range(torch.cuda.device_count()):
            devices.append(f"{torch.cuda.get_device_name(i)}:{i}")
    devices.append("cpu")     < ------- add cpu in the end 
    return devices

Describe alternatives you've considered
For now I manaually do that in my init of my infer as

class myNetwork(BasicInferTask):
    def __init__(self,path,network,
        parameters,
        type=InferType.SEGMENTATION,labels=None,dimension=3,
        **kwargs,
    ):
        super().__init__(path=path,network=network,type=type,
                    labels=labels,dimension=dimension,description=description,
                    **kwargs,)

        self._config["device"]= device_list()+["cpu"]  ## adding cpu to use 

Start in monailabel/utils/others/generic.py around lines 192-200 and inspect device_list(). Verify how the returned devices are used by the inference configuration. The change is complete when CPU can be selected alongside available GPUs and remains available when CUDA is unavailable.