ci: Capture overhead in node app

[mydea]

This adds a simple node express app that is run on CI to get very ~~ overhead measurements of using Sentry.

THIS IS NOT NECESSARILY REPRESENTATIVE OF USING SENTRY IN A REAL APP! The app is very synthetic and real world numbers may vary wildly.

We run 3 endpoints with 3 scenarios each:

1. Minimal GET request:

app.get('/test-get', function (req, res) {
  res.send({ version: 'v1' });
});

2. POST request with a bit of compute work:

app.post('/test-post', function (req, res) {
  const body = req.body;
  res.send(generateResponse(body));
});

function generateResponse(body) {
  const bodyStr = JSON.stringify(body);
  const RES_BODY_SIZE = 10000;

  const bodyLen = bodyStr.length;
  let resBody = '';
  for (let i = 0; i < RES_BODY_SIZE; i++) {
    resBody += `${i}${bodyStr[i % bodyLen]}-`;
  }
  return { version: 'v1', length: bodyLen, resBody };
}

3. GET request with MySQL query:

app.get('/test-mysql', function (_req, res) {
  pool.query('SELECT * from users').then(([users]) => {
    res.send({ version: 'v1', users });
  });
});

Each of these is run for 10s (after warmup) via autocannon, and we capture the average # of requests per seconds. The scenarios are:

1. Plain, no Sentry/instrumentation
2. With "full" Sentry instrumentation:

Sentry.init({
  dsn: process.env.E2E_TEST_DSN,
  tracesSampleRate: 1,
});

3, With errors-only Sentry instrumentation:

Sentry.init({
  dsn: process.env.E2E_TEST_DSN
});

[github-actions]

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,004	-	-	added
GET With Sentry	1,317	15%	-	added
GET With Sentry (error only)	4,175	46%	-	added
POST Baseline	1,196	-	-	added
POST With Sentry	496	41%	-	added
POST With Sentry (error only)	944	79%	-	added
MYSQL Baseline	3,306	-	-	added
MYSQL With Sentry	419	13%	-	added
MYSQL With Sentry (error only)	2,228	67%	-	added

[Lms24]

Nice, thans for adding this!

(as discussed, some more involved endpoints would also be interesting to test against but we (anyone in the team) can add these as a follow-up)