Specify 'logs root' property of L2 Blocks

[maurelian]

We should consider including a "logs root" (something more neat than the receipt root on L1).
Merkleizing the L1 log events more efficiently, filtered to only those of the successful receipts, would be a nice thing to do as rollup. If L2 doesn't pay for a log-subscription in advance (the extension we talked about), at least it can efficiently provide a proof later on that something happened on L1.

This issue is a placeholder, taken from a PR comment here by @protolambda.

[norswap]

Is this worth it? The merkleized entries in the receipts root comprise a concatenation of all log events emitted during a transaction. Seems like except extremely weird cases, this should not be too big? EIP-4488 will also drastically reduce the cost of proving the entry.

[protolambda]

It's not about the cost of the fraud-proof here, but about the cost for a L2 user inside the L2 EVM to casually proof some L1 event.

If we did the heavy lifting by providing a nice to work with accumulator then L2 users will have to pay less fees and it may also be less error prone for them (a lot easier to verify a plain byte array of a log entry in a binary tree, than to verify a RLP-encoded receipt in a MPT)