This can be used to disable vectorization and memcpy/memset
expansion for things like OS kernels. It also disables implicit
uses of scalar FP, but I don't know if we have any of those for
RISC-V.
NOTE: Without this patch you can still do -Xclang -no-implicit-float