try to fix a bug in proving

[dantaik]

@adaki2004 There are bugs in LibProving. Tests will fail for the following scenario on main:

1. Assume realProofSkipSize is 0 or 1.
2. Submit an oracle proof
3. Submit a real proof to override the oracle proof

I tried to fix the bug but don't have much time today, so please take it as priority. I'd like to define the following terms to help you understand things:

• An oracle proof: if evidence.prover is address(0), the proof is an oracle proof. On main branch, we sometimes override evidence.prover to a non-zero value, this is wrong, we should not do that.
• A fake proof: if the evidence.prover is the oracle prover address, then this proof can be called a fake proof. Fake proofs do not run verifiers, besides that, they are the same as regular proofs.
• A regular proof: if evidence.prover is not 0 or the oracle prover address.

We can have two different addresses for "oracle_prover" and "fake_prover" respectively, then things will be easier. In the code, we just use the same address for both oracle proofs and fake proofs.