Content-Length invalid for unicode text

When generating a http request string, the content-length seems to be calculated using the number of codepoints instead of bytes.
This was generated using insomnia, which uses httpsnippet:
```
POST /api/v1/test HTTP/1.1
Content-Type: text/plain
User-Agent: insomnia/2023.5.8
Host: localhost:5003
Content-Length: 3

äöü
```

This is what insomnia actually sends when executing the request (visualized using nc -l localhost 5003)
```
$ nc -l localhost 5003
POST /api/v1/test HTTP/1.1
Host: localhost:5003
Content-Type: text/plain
User-Agent: insomnia/2023.5.8
Accept: */*
Content-Length: 6

äöü
```

Start by tracing the code that generates the HTTP request string and calculates Content-Length, then check how Unicode text is encoded. Add a regression test using äöü and verify that the generated length matches the UTF-8 byte count.

Vom Indexierungsmodell aus dem Issue-Text verfasst.