Puppet agent should not fall back to puppet:0 when servers in server_list are unreachable

Repro steps:

• Install a puppet master
• Remove 'server' setting from puppet.conf
• Add 'server_list' setting to 'puppet.conf' with the master hostname and a port that is unreachable - {hostname}

  :6666

• Run puppet agent -t --debug

Result:

• You see the following in stdout:

Debug: Creating new connection for https://rjvrvpaz5nkmddc.delivery.puppetlabs.net:6666

Debug: Using cached certificate_revocation_list for ca

Debug: Starting connection for https://rjvrvpaz5nkmddc.delivery.puppetlabs.net:6666

Warning: Could not select a functional puppet master

Debug: Failed to load library 'msgpack' for feature 'msgpack': cannot load such file -- msgpack

Debug: Puppet::Network::Format[msgpack]: feature msgpack is missing

Debug: node supports formats: json pson yaml

Debug: Creating new connection for *https://puppet:0*

Debug: Starting connection for *https://puppet:0*

Trying to run against https://puppet:0 takes a while to time out due to TCP timeout (since DNS lookup fails). Also, not sure of the impact but someone could potentially set up a server named puppet on the network and intercept all requests intended for the master in the case where the master is unavailable.

Expected:
We should not be falling back to puppet:0