Single-row Optimizations on top of KeccakRound gate

[querolita]

These optimizations still keep the KeccakRound gate in one single row. Other optimizations will build on top of this one to split across rows to create narrower chips. Here, by "Old" I mean the design shown in the RFC.

• Remove intermediate states and embed them in constraints
• Simplify rotation inside Theta removing bound_c and lookups for quotient_c (which just needs a boolean check)
• Remove bound_e following the discussion about optimizing rotation with the product team

Old column count: 2344
New column count: 1965

Old constraint count: 754
New constraint count: 389

Old lookup count: 1760
New lookup count: 1620

Old prover time: 23s x 1 hash | 1871s x 1000 hash
New prover time: 19s x 1 hash | 1691s x 1000 hash

Old setup time: 57s x 1 hash | 517s x 1000 hash
New setup time: 57s x 1 hash | 479s x 1000 hash

Old verifier time: 6516ms x 1 hash | 6797ms x 1000 hash
New verifier time: 5600ms x 1 hash | 6021ms x 1000 hash

[querolita]

Making it a draft for now. The gate changes are in #1542. Will reopen when the proof system is modified to be compatible with these branches.

[querolita]

Closing because the constraints and witness have already been merged, and we probably won't be integrating kimchi's proof system with these for now