According to this, a HTTP response's header is separated from the body using CRLFCRLF.
However, reading through rfc9110, I did not find anything about that.
Where is it standardized that this should be the case?
If it is not, how is one supposed to extract the body of an entire HTTP response "safely"?
RFC 9110 defines those parts of HTTP which are the same for all versions of it.
What you are looking for is an aspect of HTTP/1.1, and that is defined in RFC 9112.