improve rrot7 for u8 and u32

[Hakkush-07]

This PR improves u32_rrot(i) for i=7, 15, 23, 31 by improving u8_rrot7, u8_extract_1bit, and u32_rrot7 functions.

u32 rrot 7 costs 76 instead of 86
u32 rrot 15 costs 77 instead of 85
u32 rrot 23 costs 77 instead of 87
u32 rrot 31 costs 78 instead of 86

This slightly reduces sha256 script size.