Reshape creates a copy of the data but shouldn't

[tgkolda]

The method tensor.reshape should not make a deep copy of the data. Either reshape should change the tensor it's called on, or it should be a function rather than a method. In numpy, the default behavior of reshape is that it returns a view (if possible, depending on the order). I think we get in trouble because we then call the tensor constructor which makes a copy. We need an option to not make a copy. Ideally, we can have two tensors with different shapes that share the same underlying data, but this needs to be clear so that the user knows when they have a view versus a copy.