[Feature Request]: implement stubs for BadEncodingFraudProofs

[adlerjohn]

Summary

Tracking issue: #4

Implement stubs to produce and verify BadEncodingFraudProofs.

Impact on users

BadEncodingFraudProofs are necessary to have secure partial nodes.

Cost of not doing

No partial nodes ☹️

Evidence

No response

Implementation ideas

As the first step towards having BadEncodingFraudProofs, add stubs to the node for functions to produce and verify BadEncodingFraudProofs.

Producing BEFPs

Full node starts performing FullAvailability check.

celestia-node/node/services/service.go

Line 137 in 40c4c62

return share.NewFullAvailability(merkledag.NewSession(fxutil.WithLifecycle(ctx, lc), dag))

During FullAvailability, repairing the square is attempted.

celestia-node/ipld/read.go

Lines 60 to 62 in 40c4c62

	// repair the square
	tree := wrapper.NewErasuredNamespacedMerkleTree(uint64(edsWidth) / 2)
	return rsmt2d.RepairExtendedDataSquare(rowRoots, colRoots, flattened, codec, tree.Constructor)

The error is returned immediately and not processed. RepairExtendedDataSquare needs to be changed to check that the error is not ErrUnrepairableDataSquare (which indicates a need for more samples, orthogonally), and if not, instead create a BEFP using

1. The returned ErrByzantineRow or ErrByzantineCol, and,
2. A Merkle proof for shares in the above error that are missing from the error struct.

The second point will require retrieveShare to return a Merkle proof along with the share value.

celestia-node/ipld/read.go

Line 49 in 40c4c62

go sc.retrieveShare(rootCid, true, row, col, dag)

A produced BEFP must then be disseminated across the network with some new p2p message (a separate issue).

Verifying BEFPs

When receiving a BEFP from the network (a separate issue), it is verified.

Urgency

Medium

[renaynay]

Closing in favour of #528