[RFC][SPIR-V] Way to represent float8 in LLVM IR

Code Generation
,
10

@arsenm @nikic thanks for the feedback! I’ll draft the patch defining the intrinsics then.

For the builtins - we still need them for SPIR-V consumers, that are based on not-ToT LLVM, but they will be kept internal for SPIR-V to LLVM IR translator.