Add LTO + PGO + BOLT optimizations

Hi!

For gaining even more performance I suggest trying to integrate into the Codon LTO ("fat" version) + PGO + LLVM BOLT set of optimizations. This set is already used by compilers like `rustc` and could be helpful for Codon too. More about PGO results for real-life applications you can find here - https://github.com/zamazan4ik/awesome-pgo .