intel/vec4: Refactor operand fixing for ffma and flrp