try adding parallelisation

[alxvov]

1. I'm using simple parallelisation of the outer loop j, which first should parallelise over columns for XOR transform, and second should parallelise over rows for Hadamard transform. Since Hadamard applied to each row independently, parallelisation should work well.
2. setup.py needs work.
3. If anyone knows how to do it better, please comment.

EDIT: seems parallelisation of XOR over second for loop works better