Optimize bn128 precompiled contracts

[AllFi]

What does this PR do?
This PR optimizes BN128Addition, BN128Multiplication, and BN128Pairing precompiled contracts by using arkworks implementation of bn128 operations. This PR uses LibarkworksWrapper introduced in zkBob/zksnark-java-sdk#1.

Why are these changes required?

The native implementation of these precompiles is relatively slow. In general, this hinders on-chain zkSNARKs verification and, consequently, makes zk-based apps almost unfeasible (tronprotocol#4311). With this PR (or a similar approach), these applications become viable.

It is important to note that these changes are not propose to solve a specific problem of ZkBob but rather enable a whole class of applications and protocols based on elliptic curve cryptography and pairings in particular. The possibilities brought by ZKP are not limited to privacy focused protocols , some other examples are zk based auth for AA wallets, light clients, zk based bridges, computational integrity enforcing contracts etc

This PR has been tested by:

• Unit Tests
• Manual Testing

Follow up

Extra details
I've added some benchmarks before (https://github.com/zkBob/java-tron/tree/bn128-native-bench) and after modifications (https://github.com/zkBob/java-tron/tree/bn128-arkworks-bench). The results can be found below.

The average time of operations before (Intel(R) Core(TM) i7-10750H CPU, 32 GB RAM):

BN128Addition: 66387 ns
BN128Multiplication: 3553350 ns
BN128Pairing (10 pairs): 101565419 ns

The average time of operations after:

BN128Addition: 11576 ns
BN128Multiplication: 181301 ns
BN128Pairing (10 pairs): 3285601 ns

[EvgenKor]

lgtm