fix(large-response-middleware): use Buffer.byteLength for accurate UTF-8 size measurement

[paulofph]

Summary

• Replace payload.length * 1.0 with Buffer.byteLength(payload, 'utf8') for accurate byte counting
• Add node to tsconfig.json types so Buffer resolves without a TS error

Problem

String.prototype.length counts JavaScript UTF-16 code units, not bytes. For responses containing multi-byte UTF-8 characters the middleware was significantly underestimating the actual payload size, causing it to miss the error threshold and pass an oversized response to the Lambda runtime — which then hard-rejected it with:

LAMBDA_RUNTIME Failed to post handler success response. Http response code: 413.
{"errorMessage":"Exceeded maximum allowed payload size (6291556 bytes)."}

This meant the client received a 413 even when it had correctly sent Accept: application/large-response.vnd+json.

Test plan

• All 8 existing tests pass (npm test)
• tsc --noEmit clean

🤖 Generated with Claude Sonnet 4.6

[alexmarqs]

lgtm! thanks 🤝