Use zkevm-sha256 gates for Sha256Wide chip

[nulltea]

Advantages:

• more secure, written not by us (source in pse/zkevm-circuis and axiom/halo2lib).
• no RLC, thus no need to generate secure randomness
• significantly less columns, almost the same height
• much more efficient (see benchmarks below)

Benchmarks

Config	Circuit degree	Prover time	MSMs in 1 stage	Contract size	Gas cost
Old	20 (1 stage), 25 (2 stage)	220s + 610s = 830s	255	?	?
New1	20, 25	155s + 475s = 630s	157	23552	349791
New2	20, 24	155s + 310s = 465s	157	26267	fail to deploy
New3	19, 25	80s + 460s = 540s	168	23552	349779
New4	18, 25	45s + 460s = 505s	189	23550	349839

• first stage 2,972,788 advice cells vs 3,729,373 (old)
• peak memory usage (new4) - 106GB

[ec2]

Good stuff. Love having less code