Price delayed publications for proving and modify the ProverMarket

[ggonzalez94]

As part of the ahead of time auction design, we have the idea of pricing delayed publications differently.
If we go down this path, calculating rewards when there are delayed publications becomes more tricky, at least with the current proposed implementation. This is because we are currently not storing in the Period struct wether a publication is delayed or not.
Once we decide if we want to price them differently, we should implement a mechanism to do so in the ProverManager

[ggonzalez94]

This is our thinking so far(I'll try to create a PR for this soon)

• We'd like to provide the support for pricing delayed publications differently(they may be bigger and also you have time uncertainty on what the price will be by the time they are included)
• Whatever multiplier you choose, you can never be sure it will be enough to cover the actual proving cost when the delayed publication is included since the bids may have changed significantly(of course the shorter the delayed inclusion period is and the higher the multiplier you can be very confident)
• In the unlikely case that the payment of the delayed publication wasn't enough we think the best tradeoff is asking the proposer to cover the difference. The other option is to ignore the publication, and since this is an escape hatch mechanism we want to avoid that since it can become a censorship mechanism
• We don't want to store individual fees per publication(given storage costs of doing so). A solution for this is to add the fees for the delayed publications directly to the stake of the period. This way the current prover can always reclaim them when they finalize their period. The only case where this wouldn't work is when they are evicted. In this case the accumulated fees will be divided between evictor, new prover and burnt. While not ideal, we believe this is an ok trade-off.

[ggonzalez94]

After starting to work on this I realized we still needed the prover to prove how many were non delayed publications(since we were using that for calculating the fees). Because of that I believe it makes more sense to treat the delayed publications as part of the fees, and just differentiate them when calling prove