fix: 🐛 request.url getter fails when using IPv6 and HTTP2 [#4560]

[gsoldevila]

Summary

Fixes #4560

Problem

In HTTP2 the Host: header is no longer present. According to the spec:

In HTTP/2, the functionality of the HTTP/1.1 Host header is replaced by the :authority pseudo-header, which is a mandatory, colon-prefixed header that specifies the authority (host and port) of the target server. This change is part of HTTP/2's binary framing and header compression (HPACK), and any HTTP/2 request converted to HTTP/1.1 must create a Host header from the :authority pseudo-header.

• When building the request object and extracting its information (here), we rely solely on Host: header.
• Then, when starting Hapi in HTTP2 mode and performing a request, the Host: header is missing, so the request.url getter) uses the host:port from the _core.info.
• Now, if the server is configured with an IPv6 host (such as ::1), the getter ends up building an invalid URL, as the host name is not surrounded by square brackets. An IPv6 address like 2001:db8::1:8080 would be ambiguous, as 8080 could be interpreted as the last segment of the IP address rather than the port.

Solution

The PR updates the request logic in 2 places:

• First, we default to using :authority if Host header is NOT present.
• Then, if for any reason both headers are missing and we default to using the server info, we make the logic support the IPv6 scenario by wrapping the hostname (e.g. ::1) in square brackets [::1].

[gsoldevila]

Perhaps we could test if the req is an IPv6 request somehow, and choose Host or :authority accordingly. WDYT?