Implementation of new MIL algorithms and editing existing files to accommodate such algorithms

[awxlong]

1. I first deleted MIL.py from PL_protocol/ and replaced it with my own BaseMIL.py. This implements the abstract class BaseMIL from which all MIL algorithms inherit.
2. I modified Image/MIL/init.py's list of available MIL models. I also modified Image/MIL/utils.py's file to include more helper functions relevant to the MIL algorithms, such as transformer layers and multimodal fusion.
3. I modified the Image/init.py create_img_model() function to be more flexible to call the newly implemented MIL algorithms
4. I added the definition of the following MIL algorithms, their paras, and pytorch lightning wrapper: AttentionMIL, CAMIL, CLAM, DTFD_MIL, DTFDMultimodal (please ignore this one I shouldn't push this), GraphTransformer, Transformer, TransMIL, and TransMILMultimodal and TransMILRegression.