Use /LTCG and /GL on and their GCC counterparts (Maybe even do PGO) - up to 40% performance boost overall

[Pirulax]

/LTCG - Link-time code generation
/GL - Whole program optimization

S.O. answer about /LTCG (/LTO on GCC)

Why? Free performance from what I can see. I compiled it on Windows, and got a nice 10-30% improvement in speed for my event handler stuff (#2173).
I dont think we'll have any issues regarding static obj init, as we aren't using global static objects at all. (Or not in a way that they depend on each other).

Also, we could make use of PGO on Windows (Pretty sure there's a Linux counterpart, but server performance is already good, it's the client which needs a lot of improvements.)
It seems like it can speed up the Lua interpreter by about 30% (Source).

Theres a lot of free performance on the table IMHO, and it seems like it's pretty trivial to collect it.
I'm pretty sure /LTCG /GL can net us 10-20% + PGO another 20-30%