fix: move res.end() outside stream reading loop in setResponse

[roor0]

Problem

setResponse() in the Node.js adapter calls res.end() inside the stream reading loop (after a successful res.write()), causing the HTTP response to be terminated after the first chunk (~16KB). Any response body larger than one chunk is silently truncated, returning incomplete/malformed JSON to the client.

This affects applications using better-auth's customSession plugin where the session payload exceeds 16KB (common with production user data containing assignments, projects, divisions, etc.). The client receives truncated JSON, parses it as null, and the user appears unauthenticated.

Root Cause

// Before (broken):
async function next() {
    try {
        for (;;) {
            const { done, value } = await reader.read();
            if (done) break;
            if (!res.write(value)) {
                // backpressure handling...
            }
            res.end(); // ← called inside loop, ends response after first chunk
        }
    } catch (error) { ... }
}

Fix

Move res.end() after the loop:

// After (fixed):
async function next() {
    try {
        for (;;) {
            const { done, value } = await reader.read();
            if (done) break;
            if (!res.write(value)) {
                // backpressure handling...
            }
        }
        res.end(); // ← called after loop, full response is written
    } catch (error) { ... }
}

Test

Added a test that creates a response body larger than 16KB and verifies the full body is written to the Node.js response without truncation.

[roor0]

Closing — this is a duplicate of #124 which already addresses the same res.end() placement bug. Thanks!