Enable LRO for TCP per default in the network drivers.

Enable LRO for TCP per default in the network drivers.

Large Receive Offload allows to receive aggregated packets larger
than the MTU.  Receiving TCP streams becomes much faster.  As the
network hardware is not aware whether a packet is received locally
or to be forwarded, everything is aggregated.  In case of forwarding
it is split on output to packets not larger than the original
packets.  So path MTU discovery should still work.  If the outgoing
interface supports TSO, the packet is chopped in hardware by TCP
Segmentation Offload.

Currently only ix(4) and lo(4) devices support LRO, and ix(4) is
limited to IPv4 and hardware newer than the old 82598 model.  If
the interface is added to a tpmr(4), bridge(4) or veb(4), LRO is
automatically disabled.  All ix(4) devices support outgoing TSO for
IPv4 and IPv6.  Enabling LRO on lo(4) automatically enables TSO and
TCP packets larger than the MTU pass the loopback interface.

LRO can be turned off per interface with ifconfig -tcplro.

OK jan@