arm: Add VFP-accelerated version of int32_to_float_fmul_scalar