[kanchingyg]

Fixes #183

readStream in src/main.ts converts each stream chunk with chunk.toString(). When a subprocess writes UTF-8 output and a 2-, 3-, or 4-byte character is split across two Buffer chunks, each half is decoded in isolation and replaced with U+FFFD.

This change uses StringDecoder('utf8') from node:string_decoder in readStream (decoder.write(chunk) per chunk followed by decoder.end()) without mutating the stream encoding on this.process.stdout.

Added an async test in src/test/main_test.ts that synchronizes two stdout writes split mid-character (buf.subarray(0, 3) then buf.subarray(3)) and verifies the decoded output matches 'ñé€🚀'.