CIFAR-10 dataset in pipeline

[philswatton]

• add CIFAR-10 to pipeline
• add generating target-surrogate pairs to pipeline
  • for targets, subset from CIFAR-10
  • for surrogates, apply transformations
    • removal of examples
    • data augmentation
    • both together

[philswatton]

First step: establishing the simplest dataset modifications. Where A is the target dataset and B is the surrogate dataset:

• Create A, copy into B, remove examples from B
• Create A, copy into B, remove examples from A

Once this is done, work out:

• Create A, copy into B, remove (non-overlapping) examples from A and B

[philswatton]

#9 implements the simple case of removing examples from one but not both of A and B

[philswatton]

With #30 done, this issue is now complete