[clang-tools-extra] Bfi precision (PR #66285)

[Duncan P. N. Exon Smith via cfe-commits]

dexonsmith wrote:

Seems awkward to pessimize as "cold" when there's no real data (no PGO, no other marking). What happens if you change the cold call threshold to 0% when there's no PGO data? (I.e., never assume a call is cold without actual evidence)

https://github.com/llvm/llvm-project/pull/66285