* optimize v_lut_deinterleave
 * optimize v_interleave_/pairs/quads/triplets
 * optimize v_lut, use vec_extract instead of aligned store