Switch to CMake for build process

[GUI]

I recently stumbled upon CMake's ExternalProject functionality. I hadn't really used CMake before, but the ExternalProject feature solves our need to build other project dependencies within our project, and then bundle them all together (for example, OpenResty, Ruby, TrafficServer, etc). We had devised our own solution to this in our custom Makefile, but it had gotten a little unwieldy, and CMake solves this much more cleanly and robustly (and eliminates issues like build caching issues, we've tried to hack around).

So while I like the simplicity of standard Makefiles, and I'm not super-fond of switching our build process on a whim, there are several drivers for this that seem to make it worthwhile:

• The aforementioned ExternalProject eliminates a lot of our custom code that's not nearly as robust.
• After working to switch things over, it cuts the size of our build code in half (and the CMake version could still probably be cleaned up and made more modular).
• CMake also provides CPack, which can generate binary packages (RPMs, DEBs) from the build. This more integrated approach helps cleanup and standardize our packaging process, and eliminates the need be using FPM to do this.
• The primary driver for this switch happening now is that with CPack we can more easily generate multiple packages from the same build. I want this so that we can ship an optional api-umbrella-hadoop-analytics binary package as a way to package up the new hadoop-based analytics work, but in an optional fashion (so people not using hadoop for analytic won't have to have all those extra dependencies). It would be possible to build a separate package with our current Makefile & FPM approach, but CPack seems to over a much cleaner and standardized way to do this, rather than us inventing our own solution.
• While CMake isn't quite as standard as standalone Makefiles, CMake's still a pretty common build tool that's widely available (and ends up generating makefiles), so it doesn't seem that different.

There's still some loose ends to tie up with the CMake functionality in this branch, but since one of the primary things I want from this switch is the ability to generate the separate "hadoop-analytics" package, I'm going to go ahead and merge this into the analytics-revamp branch to finalize some of those packaging details there. So this change won't really be going live until we fully wrap-up and ship the analytics branch, but I wanted to open a separate pull request to delineate this functionality and see if anyone has any opinions on CMake (we can always back out of this change on the branch later if anyone is opposed).