feat(protocol): partially randomize prover reward

[dantaik]

Introduce a new config: proverRewardRandomizedPercentage. If it is set to 0, then the proof rewards are distributed as before this PR, if it is 100, then the proof rewards are distributed randomly. We can choose a value between 0 and 100 so the reward are partially random and partially fixed.

[Brechtpd]

I have some doubts about the PoWification of proof generation.

• Rewards less efficient provers which could be good, but it seems inevitable that these will still end up being unprofitable because the proof price will adjust down until the price where the more efficient provers are just profitable. So seems likely to me that these provers will still not want to prove blocks because it will unlikely to be profitable for them.
• The above could be untrue if the protocol knows it overpays for the proof generation, and faster provers cannot take advantage of these higher price. But higher prover costs will increase transaction costs so not great for users and and a potential disadvantage compared to other rollups. But that may also not true because
• The protocol doesn't know if indeed different provers are submitting proofs. You would think that the most efficient prover submits all the proofs until it's not profitable anymore, and so less efficient provers which would be good for "decentralization" are still out-competed. But the extra transaction fees paid by the users actually did not help get multiple different provers to proof the block, so the users pay a higher transaction fee for no benefit at all.
• It actually seems good to me that inefficient provers are not profitable and will stop generating proofs? Does multiple provers proving the same block really benefit users?
• The PoW factor also seems to make uncle rewards unnecessary because they have the same goal with a different implementation. Which is actually a good thing because that means always only a single proof to check on-chain.

I'm not sure it's possible to come up with a good mechanism that achieves the specified goal of actual multiple distinct provers proving the same block without knowing the identity of provers.