fix(index): decode child stdout/stderr as one contiguous buffer

[Fdawgs]

Crazily I was just looking to reduce the transient heap size that would bloat during conversion but ended up finding a chunk boundary bug.

This PR updates convert() to collect stdout/stderr chunks into Buffer arrays and decode once on close via Buffer.concat().toString() instead of accumulating them with stdOut += data.

The previous approach implicitly called Buffer.prototype.toString() per chunk, which decoded each chunk independently. Multibyte UTF-8 sequences split across pipe chunk boundaries were replaced with U+FFFD, silently corrupting output.

Decoding once over contiguous bytes fixes the boundary bug and lowers peak heap use during conversion by ~40-50% in benchmarks.

Checklist

• Run npm test
• Commit message adheres to the Conventional commits style, following the @commitlint/config-conventional config

[Copilot]

Pull request overview

Fixes UnRTF#convert() output corruption when UTF-8 multibyte sequences are split across stdout/stderr chunk boundaries by buffering raw bytes and decoding once at process close, which also reduces transient heap pressure during conversion.

Changes:

• Update convert() to accumulate stdout/stderr as Buffer[] and decode once via Buffer.concat(...).toString("utf8").
• Add unit tests that simulate chunk-boundary splits for 2-, 3-, and 4-byte UTF-8 characters to prevent regression.
• Minor test refactor to reuse EventEmitter/Readable imports for child process mocks.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/index.js	Buffers child process output as contiguous bytes and decodes once to avoid UTF-8 boundary corruption.
test/index.test.js	Adds regression tests for UTF-8 chunk-boundary splits and tidies child_process mocking setup.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.