intel: Fix extract_i8 and extract_u8 for 64-bit integers
I'd really appreciate review from @jekstrand and @mattst88 on "intel/fs: Fix extract_u8 of an odd byte from a 64-bit integer" and "intel/fs: nir_op_extract_i8 extracts a byte, not a word" since we had discussed the issue on IRC so much.
The nir/algebraic patches don't help anything in shader-db, so I won't be opposed to dropping them. They did, however, generate some of the patterns that the old brw_fs_nir code did not handle correctly, and led to failures in the recently added piglit test cases.
Edited by Ian Romanick