[P3][adapters] Read WAL by fd/offset in node-fs instead of whole-file readFileSync

[cevheri]

Summary

node-fs adapter reads via readFileSync(path) on every read() call — whole file into memory. At open(), recovery reads entire WAL → transient 2× memory. Uses path not fd — wrong inode if path swapped mid-session (edge case).

Audit ID: Section 4 Low — Wave 3
Related: #12 compaction (large files amplify cost)

Fix direction

• Read via fd + readSync at offset/length.
• Optional: mmap-style single buffer cached on WalFile for session.

Acceptance criteria

• read(offset, length) does not load entire file when length << size.
• Tests with spy/mock or large temp file.
• Changeset if performance observable.

Agent constraints

• Keep sync API; match kernel WalFile contract.

[cevheri]

Implemented and shipped in PR #43 (merge commit 69f7161), released as v0.2.0. The PR description's comma-separated Closes list only auto-closed the first reference — closing manually. For #39/#28-style partials: the shipped scope is done; the real-browser verification half of #39 continues under #10, and the typed-error contract (#28) shipped as LibreDbError with stable codes across kernel, adapters, lenses, and CLI.