fix(trie): do not reveal same node twice in sparse trie

[shekhirin]

Problem

The issue is that it's possible to reveal the same leaf twice if we request two proofs with two leaves that have the same prefix.

For example, in DB we have a leaf at path 0x0d07. First, we request a proof for leaf 0xd7b6990105719101dabeb77144f2a3385c8033acd3af97e9423a695e81ad1eb5. Then, we request another proof for leaf 0xd7999c7eedf8fe42767a51577f99fda631d84616191df1557bdaf61266d16a0f.

In both cases, the multiproof will have a leaf at path 0xd07, and we will try to reveal it. Current prevention for not revealing it second time is this check

reth/crates/trie/sparse/src/trie.rs

Lines 314 to 317 in a7f895e

	// If the node is already revealed and it's not a hash node, do nothing.
	if self.nodes.get(&path).is_some_and(|node| !node.is_hash()) {
	return Ok(())
	}

, but it will not pass if the first leaf 0xd7b6990105719101dabeb77144f2a3385c8033acd3af97e9423a695e81ad1eb5 was deleted after it was revealed, so we will reveal the leaf node agian.

It's problematic because we could have already deleted a leaf node, and we should not insert it again on proof revealing.

Solution

Maintain a list of revealed paths per account trie and each of storage tries.

[rkrasiuk]

should we keep these on trie level instead?