`_write()` bypasses chunk batching in chunked mode

[VikramAditya33]

UwsResponse._write() sends each chunk to uWS individually instead of using the existing writeChunk() batching system. For readable.pipe(res) with many small chunks (e.g., 4KB from fs.createReadStream), this results in hundreds of uWS.write() syscalls instead of ~4-16 batched writes.

Root cause: _write() calls streamChunk() directly:

_write(chunk, encoding, callback) {
    const buffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk, encoding);
    this.streamChunk(buffer, this.contentLengthTotal)
        .then(() => callback())
        .catch((error) => callback(error));
}

streamChunk() calls this.atomic(() => this.uwsRes.write(buffer)) per chunk with no batching.

Existing batching: writeChunk() already implements intelligent batching:

• Accumulates chunks in pendingChunks[]
• Flushes when HIGH_WATERMARK (128KB) reached
• Flushes on FLUSH_INTERVAL (50ms) timeout
• First chunk sent immediately for low latency