Blocked iteration

Created by: vchuravy

fixes #22 (closed) @mwarusz thank you for the performance analysis.

On the V100 I am running on:

Kernel	Time	Speed of Light Mem %
naive (32, 32)	1.19ms	65.06%
blocked	1.20ms	64.38 %
naive (1024, 1)	3.67 ms	26.84 %
naive (1024, 1) Const	1.79ms	56.13 %
naive (1, 1024)	3.66ms	49.53 %
naive (1, 1024) Const	3.03ms	60.02 %