feat: add raw checkpoint creation

[gitferry]

This PR fixes BM-47.

This PR implements handlers upon entering an epoch boundary. In particular,

1. at the beginning of the second block after a new epoch begins, a raw checkpoint should be created according to the epoch number and the LastCommitHash extracted from the block header.
2. The newly generated checkpoint's status is set to be ACCUMULATING.

The following PR will be implementing the BLS signer which creates a BLS-sig transaction and distributes it to the network.

[aakoshh]

Looks good, except for the boundary check which really needs encapsulation :)

[aakoshh]

This underlines the point of this conversation #32 (comment) with @SebastianElvis

It doesn't look right: GetEpochBoundary, as per the comment, returns the last block. We are looking for the 2nd block of the current epoch, so maybe this should be block.height - 2 == previous_epoch_boundary.

This is why that Epoch interface would be cool, you could add a method to ask IsSecondBlock(ctx) and it would figure out what you mean. Currently if we have epochs 1000 blocks long and we are in block 1002, this would compare that against 2000 I think.

[gitferry]

Yeah, I agree that the Epoch interface is cool.

It doesn't look right: GetEpochBoundary, as per the comment, returns the last block. We are looking for the 2nd block of the current epoch, so maybe this should be block.height - 2 == previous_epoch_boundary.
This is why that Epoch interface would be cool, you could add a method to ask IsSecondBlock(ctx) and it would figure out what you mean. Currently if we have epochs 1000 blocks long and we are in block 1002, this would compare that against 2000 I think.

You are right, the current GetEpochBoundary does not give what we want. If we don't want change GetEpochBoundary, the checkpointing module can implement the AfterEpochEnds hood to store the epoch boundary of the previous epoch so that we don't need to call GetEpochBoundary during BeginBlock but obtain the epoch boundary from the storage.

Or, we can insert epochNum into GetEpochBoundary(ctx, epochNum) so that we can get the boundaries we want. What do you think?

[aakoshh]

I would implement the Epoch interface and return something that can tell us everything about the current (or historical) module, and keep using the epoching module as the source of truth, rather than all modules subscribing and storing their own pieces.

[aakoshh]

Doing so would allow us to encapsulate the logic of how the boundaries are calculated. Currently it's at equal intervals, so what we actually return could be something like type epochBoundary struct { epoch_number, epoch_length } but nobody should interact with that directly. If things change and the length becomes uneven, this can be accommodated, but the interface will not change.

[gitferry]

I see. It makes sense. I'll add a TODO here and wait for @SebastianElvis to finish the epoch interface. Thanks!

[gitferry]

Thanks for your review @aakoshh. I also fixed the duplicate prefix of the storage. Please see the updates. Thanks!

[aakoshh]

Looks good! 👍