fix(types): return false instead of panic in verify_multi_proof

[giwaov]

Summary

MerkleTree::verify_multi_proof panicked instead of returning false when the stack and path lengths were in an unexpected state after processing all proof flags.

Proof verification functions handle untrusted external data — proofs come from RPC responses and deserialized network messages. A malformed or adversarially crafted proof would crash the process rather than be rejected as invalid.

The same function already returns false in two earlier guard checks for invalid proofs:

if path_len < proof.flags.iter().filter(|&&f| !f).count() {
    tracing::debug!("invalid multiproof: too few path hashes");
    return false;
}

if leaves.len() + path_len != proof.flags.len() + 1 {
    tracing::debug!("invalid multiproof: invalid total hashes");
    return false;
}

The final match arm was inconsistent with these:

// Before
_ => panic!("invalid multiproof: invalid total hashes"),

// After
_ => {
    tracing::debug!("invalid multiproof: invalid total hashes");
    return false;
}

Test plan

• Confirm that passing a malformed MerkleMultiProof (e.g. mismatched flags/path lengths that bypass the early guards) returns false instead of panicking
• Confirm existing multi-proof tests still pass

[poszu]

I suppose the intent of the author was to show that this path is not possible to be executed (even on invalid input) but it's difficult to prove this is true.