We use these two IR ops for PSRLQ and PSLLQ respectively, Which is actually implemented in a quite inefficient way. Instead switch the ops over to using VExtr which maps directly to one instruction and can emulate both.