Got DAP-2590 working

This branch can build factory images for the D-LINK DAP-2590-A1. Basics are working. LEDs no so much.
Just use factory images for sysupgrade... I taught the scripts not to need a special binary.

The kernel patch needs a lot of work. Hopefully I got the build system about right and didn't ruin builds for the Ubiquity ER/ERLITE or generic Octeon... I think those will produce under a different filename now since I had to segregate into different targets.

https://github.com/skids/source/tree/dap2590

I'm not going to send any PRs until I get a bit of feedback. Odds are I did at least some one thing badly.

So a week or so into testing this and everything major seems to work. However I did
notice one issue and I'm interested to know if anyone can replicate it on a different
kernel-4.9 platform.

The issue is that nflog sockets from an ebtables rule will start to log nothing but
len=0 "overflow" packets if you are unlucky enough to edit ebtables rules at the
wrong time.

This can be tested using a ulogd instance and will pretty quickly trigger if you
do a while true; do (add any old rule to FORWARD and then delete it); done. However
it will crop up somewhat more sporadically with other things that may dynamically alter
ebtables rules on a system which is also using nflog on ebtables rules (specifically
this problem is proving to be a blocker for deplying ebtables-dhcpsnooping.)

I haven't managed to get this to replicate anyplace else but on this AP yet,
and I don't have many test systems to test with, so I don't know whether to
blame this on the kernel version, 64-bitness, endianness, the Octeon
arch itelf, something peculiar to this particular SOC, or whatnot. So
if anyone is inclined to try the above test, It'll help me figure out where to file a ticket
and/or to look over kernel code.