[clang] [CIR] Add floating-point type descriptors to decodeFixedType (PR #194483)

[Andy Kaylor via cfe-commits]

https://github.com/andykaylor commented:

The code change is good, but the test only covers the float type descriptor. Claude suggests __builtin_ia32_vaddph256_round for half, __builtin_ia32_cvtne2ps2bf16_128 for bfloat, and __builtin_ia32_minpd for double. It looks like quad may not be reachable until we add PowerPC support.

https://github.com/llvm/llvm-project/pull/194483