dnsdist: Fix reconnection handling

[rgacogne]

Short description

The first issue was that we closed the socket toward the backend if the first connection attempt failed, but tried to reconnect without reopening it, as reported by Aleš Rygl on the mailing-list 1.
The second issue is that we would close and reconnect the backend socket on some error codes even if the error was occurring on the health check socket, which is a separate one.
This PR also makes sure we don't try to reconnect from the healthcheck and responder threads at the same time, and that we don't accidentally start the same responder thread twice.

Checklist

I have:

• read the CONTRIBUTING.md document
• compiled this code
• tested this code
• included documentation (including possible behaviour changes)
• documented the code
• added or modified regression test(s)
• added or modified unit test(s)

[rgacogne]

OP confirms that the packages built from this branch are fixing the issue for him.