[clang] [CIR][X86]Implement handling for convert-half builtins (PR #173143)

[Priyanshu Kumar via cfe-commits]

Priyanshu3820 wrote:

> hi @andykaylor, @bcardosolopes, I have decided to handle the fp16 builtins in a separate PR(was having some issues in them so thought i'd be needing some more time debugging). This PR only contains the bf16 builtins(`cvtneps2bf16_128_mask`, `cvtneps2bf16_256_mask` and `cvtneps2bf16_512_mask`). All tests pass.

f16 halfs to float builtins are now implemented. In this PR- #173572


https://github.com/llvm/llvm-project/pull/173143