[geos-devel] Add DD math to 3.8

[Paul Ramsey]

So, the follow-on to the discussion regarding master (which is now merged) is the question for 3.8.

Even though (amazingly) the DD implementation showed no (!!) changes in regression results compared to ttmath, for 3.8 I have brought the DD implementation in as an addition to ttmath, only used in the case of big-endian architectures.

https://github.com/libgeos/geos/pull/304

This should minimize the probability of a behaviour changes for existing installations, while allowing (currently broken) big endian clients to pass regression. 

If anyone has login access to a big-endian machine, I'd love to confirm that my build script changes actually work. I'm pretty sure they do, but not 100%.

P.