fix(FileStore): deduplicate merkle leaves in readMerkleFile to preven…

[es-kt]

…t wrong leaf lookup

The JSONL merkle leaves file can accumulate duplicate CID entries when appendMerkleLeaves is called after a process restart re-syncs an overlapping range. Since getMerkleLeaf uses the CID as an array index (rows[cid]), duplicates shift all subsequent lookups — causing buildLocalProofPath to retrieve wrong leaf commitments and produce invalid merkle proofs.

After sorting by CID, deduplicate adjacent entries keeping the last occurrence of each CID.

Summary

Brief description of changes.

Changes

Testing

• pnpm run test passes
• pnpm run build succeeds
• pnpm run type-check passes

Related Issues

Closes #

[gemini-code-assist]

Code Review

This pull request introduces deduplication logic for entries in the FileStore to prevent incorrect leaf retrieval caused by duplicate CID entries. The reviewer suggested refactoring the manual loop implementation into a more idiomatic reduce-based approach to improve readability.

[gemini-code-assist]

The deduplication logic is correct, but it can be implemented more concisely using the reduce method. This would make the code more idiomatic and potentially easier to read for developers familiar with functional programming patterns.

Suggested change

	if (out.length > 0) {
	const deduped: typeof out = [out[0]!];
	for (let i = 1; i < out.length; i++) {
	if (out[i]!.cid === deduped[deduped.length - 1]!.cid) {
	deduped[deduped.length - 1] = out[i]!;
	} else {
	deduped.push(out[i]!);
	}
	}
	return deduped;
	}
	return undefined;
	if (out.length === 0) {
	return undefined;
	}
	return out.reduce((acc, current) => {
	const last = acc[acc.length - 1];
	if (last?.cid === current.cid) {
	acc[acc.length - 1] = current;
	} else {
	acc.push(current);
	}
	return acc;
	}, [] as typeof out);