Infer prototype while training

[karangrewal]

The CentroidContext mixin has been augmented so that now it can do either of the following during training:

1. create prototype contexts via task labels while training (just as before)
2. use a clustering technique to create prototypes in an online way while training (new!)

There is also a corresponding config file for experiments that infer prototype while training. Two additional things to note:

• The experiment/mixin attribute contexts is now a List containing a single Tensor (instead of being just a single Tensor) since lists are mutable, and we require contexts to be mutable since it is updated regularly by other functions,
• There isn't any formal documentation of the exact clustering procedure in the mixin itself, and we can redirect to an ArXiv paper that contains this info (once it becomes available).

[karangrewal]

I am closing this PR since I believe the implementation is not entirely correct, and there are currently multiple conflicts with the master branch. This does not affect results in the dendrites pre-print.