Split zokrates_bellman crate in PlonK branch

[georgwiese]

Depends on #1249

To make the Groth16 MPC trusted setup work, we need to depend on two versions of bellman_ce. I implemented this by splitting the zokrates_bellman crate.

Also, I went through the failing tests and fixed as many as possible.

This is an overview of the changes:

• zokrates_bellman: Now depends on the matter-labs version of bellman again, commented in code in the Groth16 scheme, removed PlonK
• zokrates_bellman_plonk: A new crate that depends on our fork of bellman
  • lib.rs: Is essentially a copy of zokrates_bellman/src/lib.rs, except that it depends on a different bellman crate and uses the BellmanPlonkFieldExtensions instead of BellmanFieldExtensions
  • plonk.rs: Implementation of the PlonK scheme
• zokrates_cli: Now depends also on the zokrates_bellman_plonk crate, commented in the MPC implementation of the Groth16 trusted setup, added PlonK to integration tests
• zokrates_field:
  • Now depends on both bellman crates and defines the BellmanPlonkFieldExtensions trait in addition to BellmanFieldExtensions
  • Adjusted the bellman_extensions! macro and implemented BellmanPlonkFieldExtensions for all curves that previously implemented the BellmanFieldExtensions
• zokrates_proof_systems: Fixed a bug where solidity verifiers were rendered incorrectly, triggered by a previous refactoring of the Fq2 and GAffine data types.
• Fixed a few clippy errors

The diff between zokrates_bellman/src/lib.rs and zokrates_bellman_plonk/src/lib.rs is quite small. It could be implemented using a macro (although I'm not sure in which crate it should be defined), let me know if I should do that :)