Benchmark and optimize the tx circuit

[ed255]

In https://github.com/privacy-scaling-explorations/zkevm-circuits/pull/484 we have the first implementation of the tx circuit.

For each tx it requires (due to the signature verification):

• ~205 columns
• ~170k rows

Things to do:

• [ ] We need to run benchmarks to see the cost of calculating a proof per tx.
• [ ] There are a few optimizations that can be applied
  • [x] Reduce the number of columns by using rotations https://github.com/privacy-scaling-explorations/zkevm-circuits/issues/625
  • [ ] Reduce the number of rows needed for the wrong integer to bytes conversion using a halo2wrong specialized chip for that https://github.com/privacy-scaling-explorations/zkevm-circuits/issues/624
  • [ ] Reduce the number of rows needed for an ECDSA signature verification using batch verification: https://github.com/appliedzkp/halo2wrong/pull/22