Add the support for float16

[ElaineYao]

When running VGG model provided in the repo, I get a TypeError: Unknown type <dtype: 'float16'>. For now, TensorFI supports four types - int and floats of 32 and 64, it might be good to add the support for float16.

Also, there might be a bug at line 284 in injectFault.py

else:
         raise TypeError("Unknown type" + type) 

as it'll raise TypeError: cannot concatenate 'str' and 'DType' objects

Maybe it should be written as

else:
          raise TypeError("Unknown type" + str(type))

Thanks