Fix typo and inconsistency in the roxygen2 documentation

[DerrickUnleashed]

Fixed some typos:

• desease -> disease
• class_too_idx -> class_to_idx

Fixed model documentation inconsistencies by aligning them with the documentation that was requested during previous model creations:

• DeeplabV3
• ConvNext Detection

Other model documentations can also be aligned using the same metrics. However, I am pausing for now since the current tasks are complete and I will continue this work during the GSoC period.

[cregouby]

praise thanks a lot for the typos and for the model documentation improvement.
todo there are some revertion to do on the models documentation (see inline comments)

[DerrickUnleashed]

Thanks for the inline comments. I have now run check(), and it is successful.

Some doubts arise,

• @importFrom torch nn_module is used in some models like models-deeplabv3.R however, it is not present in the other models like models-convnext-segmentation.R. Is it necessary for it to be present in all models that use torch::nn_module()? If so, I can make the necessary updates required.
• A lot of commands in the code base utilise torch::. Can we remove it since torchvision is present in torch ?
  Eg : torch::nn_module() -> nn_module().

Other Changes Incorporated:

• changed from magick_loader() to base_loader() to use the efficiency of JPEG:: and PNG::