Fix in-use VNC socket detection for qemu

The `qemu_vm` probes the availability of a free VNC port by creating
a socket and binding to it. If that succeeds, the port is considered
free.

The problem is that `find_free_ports()` in the avocado network utils
defaults to "localhost" (=127.0.0.1) for the socket address while
qemu later binds the VNC port to "0.0.0.0".

Starting with kernel 6.2.8 / 6.3.x+, the kernel behavior changed.
While the avocado code considered a port as available, later on
the start of qemu using that port failed with an "in use socket".

As qemu additionally uses SO_REUSEADDR for the socket and this allows
certain socket configuration to both bind for 0.0.0.0 and 127.0.0.1.
It's unclear if the new kernel behavior is 100% correct, but it does
not matter since the correct thing to do is to use the matching socket
address 0.0.0.0 in avocado. Broken behavior has been tested with kernel
6.2.8, 6.4.15, 6.5.5 and 6.6.13. Fix has been confirmed using the last.

May be related:
https://lore.kernel.org/netdev/[email protected]/T/#m34ca2f2765e5532243c7d20a4279baccf2ef9194
torvalds/linux@28044fc

virttest/qemu_vm.py

@@ -3025,7 +3025,7 @@ def create(self, name=None, params=None, root_dir=None,
 
             # Find available VNC port, if needed
             if params.get("display") == "vnc":
-                self.vnc_port = utils_misc.find_free_port(5900, 6900, sequent=True)
+                self.vnc_port = utils_misc.find_free_port(5900, 6900, address='0.0.0.0', sequent=True)
 
             # Find random UUID if specified 'uuid = random' in config file
             if params.get("uuid") == "random":