Shuffle single-precision (32-bit) floating-point elements in a and b using the control in imm8, Warning: the immediate shuffle value imm is given at compile-time instead of runtime.