Use custom resolver when creating the socket

[smashery]

The custom DNS resolution rules are inconsistently applied. They are respected for the purpose of choosing a communication channel (SwitchBoard::best_comm), but then when it does the resolution itself, it ignores it. One place where this occurs is within Rex::Proto::Http::Client. Setting rhost to a domain name (instead of an IP) will initially figure out the comm by doing DNS resolution using Metasploit's custom DNS resolver; but then when it goes to actually do the request over that comm channel, it eventually just calls Socket::pack_sockaddr_in with our rhost domain name, which just uses the system

Concern

This fix may have issues with Socks proxies, in which case we shouldn't be doing our own resolution... but it looks like there's already a ticket for this (#70).

Reproduction Steps

The easiest way to observe the effect of this issue is with static rules:

• Start with a clean configuration (dns reset-config)
• Delete all of metasploit's DNS configs (dns remove -i 1)
• Add the static resolver back in (dns add static)
• Remove all static entries (dns flush-static)
• Make sure it's a clean test: dns flush-cache
• Add a single static entry pointing google.com to the wrong IP (dns add-static google.com 1.2.3.4)
• irb
• send_request_cgi({'rhost'=>'google.com','rport'=>80,'url'=>'/'})
• Prior to this fix, this should actually end up reaching out to the real google server. You'll see the result's peerinfo be pointing at the real google address. The expected behaviour is that it should hang (and eventually return nil), since we overrode the IP using our custom DNS.

For a more realistic scenario, set up a DNS server with a private DNS zone. My example is *.sccm.lab. Set up an HTTP server on a host within that DNS zone (mine is mecm.sccm.lab) Ensure this DNS server is not used by the underlying OS for its DNS resolution (so dig validHost.private.zone should fail, but dig @privateDnsServer validHost.private.zone succeeds). There are probably ways to simplify the reproduction steps, but this is the use case I encountered the issue with.

• Start with a clean configuration (dns reset-config)
• Delete all of metasploit's DNS configs (dns remove -i 1)
• Remove all static entries (dns flush-static)
• Make sure it's a clean test: dns flush-cache
• Add a dns resolver to our private DNS server: dns add --rule *.sccm.lab <privateDnsServer>
• irb
• send_request_cgi({'rhost'=>'mecm.sccm.lab','rport'=>80,'url'=>'/'})
• Prior to this fix, this would fail (rex-socket/lib/rex/socket.rb:306:in pack_sockaddr_in': getaddrinfo: Name or service not known (SocketError)`). Following the fix, it should work.

Now with static entries:

• Start with a clean configuration (dns reset-config)
• Delete all of metasploit's DNS configs (dns remove -i 1)
• Add the static resolver back in (dns add static)
• Remove all static entries (dns flush-static)
• Make sure it's a clean test: dns flush-cache
• Add a single static entry pointing mecm.sccm.lab to the correct IP (dns add-static mecm.sccm.lab <the IP>)
• irb
• send_request_cgi({'rhost'=>'mecm.sccm.lab','rport'=>80,'url'=>'/'})
• Prior to this fix, this would fail. Following the fix, it should work.

[adfoster-r7]

Thanks! Just cross-referencing #71 too, as it will need to accommodate this workflow 👍

(That's not a blocker to this PR, just another thing that will need to be fixed for that other issue when it's wired up)