issues with bitFlip-tensor fault type

[nniranjhana]

Configuring tensors with the bitFlip-tensor over all operators in mnist_nn.py returns an accuracy of 10000+, when the value should be between 0 and 1.

The accuracy returned is between the range for other fault types. Will need to debug in depth to figure out where the issue is, and what causes this inflation of accuracy. Btw, we should also have a sanity check on the accuracy output.