eip7594: Remove proof parameter from recover_cells_and_kzg_proofs

[kevaundray]

This removes the proof parameter from recover_cells_and_kzg_proofs since computing individual proofs is only favourable when there are one to five cells missing. Since most clients will start recovery when half of the cells are missing, this functionality would not be useful in practice for our usecases.

We could leave it there, ignore it and have implementations decide, but note that we do need to validate the proofs are valid according to the current API, so its not free.

This means we are essentially validating the proofs twice, since a user should only pass in proofs + cells that they have passed to verify_cell_kzg_proof beforehand. It is not a security concern if these proofs are not validated since the receiver will validate them, but given that fk20 takes about 200ms to compute all of the proofs on a single thread, I don't think it makes sense to try and optimize for the seemingly rare case of one to five missing cells/proofs

[kevaundray]

One possible blocker for this PR, is the time it takes to now compute all of the proofs. Before it was only computing the missing proofs.

This is a possible blocker because it increases the CI time.

cc @hwwhww

[jtraglia]

LGTM 👍 I agree with this change. This also removes individual proof computation in recovery. In the real world, I believe it's unlikely individual proof computation will be necessary since nodes will begin recovery as soon as they have 50% of cells. And the optimized FK20 implementations (which compute all of the proofs at once) are so fast, it would only make sense to recover individual proofs if 3 or fewer cells/proofs are missing, which is unlikely.

Also, this highlighted an issue with one of the reference tests. It wasn't updated to include the proof param before, so it was throwing an exception for the wrong reason. Not a big deal, but I'm glad that's okay now.

[asn-d6]

LGTM!