[clang] [llvm] [LoongArch] Add support for half-precision floating-point type (PR #141564)

[via cfe-commits]

Ami-zhang wrote:

> LLVM uses squash and merge. So please split it into 3 PRs if you do not want to squash them.

Ok, I'll split this into 3 separate PRs.

https://github.com/llvm/llvm-project/pull/141564