swift234

[performance-inliner] Up inline threshold to 225 so that we are at least as aggressive with inlining as LLVM is.

Currently we attempt to simulate LLVM's inlining cost model. Due to the
Swift optimizer being strictly more conservative with the inlining cost
model due to swift not having as many arithmetic simplifications (for
example), the swift level cost should always be greater than or equal to
the LLVM's cost model.

Given the aforementioned paragraph, it makes sense to at least match
LLVM's cost inline threshold given we are "simulating" it.

Swift SVN r17979