This changes Premake to use link-time optimisation in release builds.
We probably don't want to patch xCode but just use flto but I wanted to try it and the "embed" trick is good for posterity.
I'm not necessarily expecting huge performance deltas, though ScriptInterface functions might get inlined more readily, which in turn might make some code faster.
Here's a quick 5min Ai-AI replayed game: the LTO seems very slightly faster overall - I'd expect slightly larger deltas on rendering since we have more out-of-line functions there.
Point is, LTO is now a rather mature technology and enabling it seems like a no-brainer on modern compilers. The situation might not be the same on older compilers, so we might want to wait on C++14/17-ready compilers?