Move and refactor the PutBlock method

[evan-forbes]

The PutBlock method was introduced as a method to Block in #178. Having PutBlock where it is now adds network dependencies to the types package, which is undesirable. The original thinking in doing so was so that erasured data could be cached instead of recomputed. While caching the erasured data is still a good idea, it was not implemented in #178 and does not necessarily require PutBlock being a method to Block.

We should probably make PutBlock a function and move it to a different package. This will require a general refactor, which includes exporting the utility function shared by PutBlock and fillDataAvailabilityHeader.

As for where to move it, per a suggestion by @liamsi, I think we should move PutBlock and it's related code to lazyledger-core/p2p/ipld.

ref: #178 (comment)

[evan-forbes]

relevant comment about the original API defined in #170 #178 (comment)

While PutLeaves was not implemented in #178, I think it should be added when we move PutBlock

[evan-forbes]

this code will also have to be moved and consolidated with PutBlock

[evan-forbes]

I'm linking this comment to remind myself to also include the suggested change into the refactor/move of PutBlock

[evan-forbes]

Another self reminder, wrap this error

[evan-forbes]

More notes:
As pointed out by @Wondertan, it's intuitive to think that calling PutBlock would also fill the data availability header, but this is not the case. In the implementation, filling the DAH and saving the block data to IPFS should be separate, but perhaps PutBlock could could also do this if not done already.

Somewhat unrelated, in #304, we're moving the ipfs API object to the block store in order to retrieve block data, and maybe we can move the PutBlock method their too.

[evan-forbes]

I think we can now close this issue now that #334 is merged. Any other opinions?

[Wondertan]

I think we can close this once PutBlock starts to return DAHeader

[evan-forbes]

So we just need to fill the DAH of the passed in *Block if not done so already? Is the goal of this to not fill the DAH twice? If so, should we also disable filling the DAH when filling the Header?

[Wondertan]

So we just need to fill the DAH of the passed in *Block if not done so already?

Ideally, I would change the interface to receive just types.Data bc that's the only thing used in PutBlock.

Is the goal of this to not fill the DAH twice?

Yeah, not to fill and recompute it twice

If so, should we also disable filling the DAH when filling the Header?

Idk yet

[Wondertan]

@evan-forbes, @liamsi is it closed?

[liamsi]

Yes, thanks @Wondertan 🙏🏼 I'm sure we can further improve the package structure a bit (similar to changes you started in #427) but the suggestions made here are already implemented.