Before this fix, a successful connection failed to reset the per IP address counter
used to count successive connection failures, that can possibly cause a host to be blocked,
when reaching max_connect_errors.

This counter is reset by function reset_host_connect_errors().
This function is called only when connect_errors is non zero,
as a performance optimization, to avoid locking the host cache for nothing.

The problem was that the connect_error counter was not returned by ip_to_hostname(),
so that reset_host_connect_errors() was never called for successful connections.

A related minor bug was fixed also: the comparison with max_connect_error was off by one,
the comparison should use greater or equal instead of strictly greater.