[1.8] network: prefer ipv4 DNS results option

[leonardo-albertovich]

This PR adds a new configuration option that can be specified either at service level or output plugin level meant to prioritize IPv4 DNS results when trying to establish a connection.

When used at SERVICE level it will set the default system wide value that will be inherited by any output plugins (unless they actively override it).

Usage examples :

[SERVICE]
    Dns.prefer_ipv4 on

Plugin level usage example :

[OUTPUT]
    Match                     *
    Name                      tcp
    Retry_Limit               false
    Host                      problematic_host   
    Net.dns.prefer_ipv4       on

[qingling128]

Just tested this. Looks like the issue still exists.

With net.connect_timeout_log_error False, we don't see connection errors.
With net.dns.prefer_ipv4 on and net.connect_timeout_log_error False removed, we still see connection errors like:

[2022/01/14 00:44:56] [error] [upstream] connection #1912 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:44:57] [error] [upstream] connection #1856 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:44:58] [error] [upstream] connection #1968 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:44:59] [error] [upstream] connection #1984 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:45:01] [error] [upstream] connection #2008 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:45:02] [error] [upstream] connection #2044 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:45:02] [error] [upstream] connection #2024 to logging.googleapis.com:443 timed out after 10 seconds
[2022/01/14 00:45:03] [error] [upstream] connection #2072 to logging.googleapis.com:443 timed out after 10 seconds

More details:
details.txt

[qingling128]

I can no longer reproduce the issue. Either I did something silly last week, and was not testing the right version of the custom build. Or there might be some changes in GCP (e.g. Logging API no longer returns IPv6, or windows 2016 suddenly supports IPv6?) over the last few days.