overhaul of the whole chi computation concept but separating the delensing and using unified memory for the GPUs
Former-commit-id: 401aabdae2f7ee44c8294841a55d9d0bfc3ff6ff