Change the `channels_first` parameter in `PyTorchObjectDetector` and `PyTorchFasterRCNN` to `True`

[f4str]

Is your feature request related to a problem? Please describe.
PyTorch models expect inputs to be provided as channels-first. The PyTorchObjectDetector, PyTorchFasterRCNN and PyTorchYolo models have a channels_first parameter which will permute the input accordingly such that it is accepted by the model. The PyTorchObjectDetector and PyTorchFasterRCNN models default this to False which is not the expected behavior.

Describe the solution you'd like
Change the channels_first parameter in PyTorchObjectDetector and PyTorchFasterRCNN to default to True instead of False. This is the expected behavior if following torchvision tutorials. Additionally, the input_shape parameter (although unused) should be adjusted accordingly.

Describe alternatives you've considered
N/A

Additional context
N/A