Enable RX checksums in hardware

[cjdoucette]

When it is available in hardware, we should enable RX checksum validation on IPv4 headers, and possibly on UDP/TCP headers.

Note that setting the RX checksum offloads doesn't make the hardware drop packets with bad checksums; it just sets bits in struct rte_mbuf that the application can then check. For the IP checksum, GK would have to check that it's an IPv4 packet and that the checksum is not bad. For L4 checksums, GK would have to check that it's TCP/UDP and that the checksum is not bad. This would potentially involve making the GK block look at the L4 header, which is does not currently do.

The RX checksum offload bits actually have four states, explained here: http://mails.dpdk.org/archives/dev/2016-October/048097.html