clean up viscy cli display

[mattersoflight]

viscy --help prints a useful and succinct help message.

But, viscy subcommand --help prints a lot of lightning CLI info that is not relevant, e.g.,

viscy preprocess --help prints:

  --lr_scheduler CONFIG | CLASS_PATH_OR_NAME | .INIT_ARG_NAME VALUE
                        One or more arguments specifying "class_path" and
                        "init_args" for any subclass of {torch.optim.lr_schedu
                        ler.LRScheduler,lightning.pytorch.cli.ReduceLROnPlatea
                        u}. (type: Union[LRScheduler, ReduceLROnPlateau],
                        known subclasses:
                        torch.optim.lr_scheduler.LRScheduler,
                        monai.optimizers.LinearLR,
                        monai.optimizers.ExponentialLR,
                        torch.optim.lr_scheduler.LambdaLR,
                        monai.optimizers.WarmupCosineSchedule,
                        torch.optim.lr_scheduler.MultiplicativeLR,
                        torch.optim.lr_scheduler.StepLR,
                        torch.optim.lr_scheduler.MultiStepLR,
                        torch.optim.lr_scheduler.ConstantLR,
                        torch.optim.lr_scheduler.LinearLR,
                        torch.optim.lr_scheduler.ExponentialLR,
                        torch.optim.lr_scheduler.SequentialLR,
                        torch.optim.lr_scheduler.PolynomialLR,
                        torch.optim.lr_scheduler.CosineAnnealingLR,
                        torch.optim.lr_scheduler.ChainedScheduler,
                        torch.optim.lr_scheduler.ReduceLROnPlateau,
                        lightning.pytorch.cli.ReduceLROnPlateau,
                        torch.optim.lr_scheduler.CyclicLR,
                        torch.optim.lr_scheduler.CosineAnnealingWarmRestarts,
                        torch.optim.lr_scheduler.OneCycleLR,
                        torch.optim.swa_utils.SWALR,
                        lightning.pytorch.cli.ReduceLROnPlateau)

Compute dataset statistics before training or testing for normalization:
  --data_path DATA_PATH
                        (required, type: <class 'Path'>)
  --channel_names CHANNEL_NAMES, --channel_names+ CHANNEL_NAMES
                        (type: Union[list[str], Literal[-1]], default: -1)
  --num_workers NUM_WORKERS
                        (type: int, default: 1)
  --block_size BLOCK_SIZE
                        (type: int, default: 32)

@ziw-liu please fix this.