AVX128: vmovq can be improved
```json
    "vmovq xmm0, qword [rax]": {
      "ExpectedInstructionCount": 4,
      "Comment": [
        "Map 1 0b01 0x6e 128-bit"
      ],
      "ExpectedArm64ASM": [
        "ldr q2, [x4]",
        "mov v16.8b, v2.8b",
        "movi v2.2d, #0x0",
        "str q2, [x28, #16]"
      ]
    },
```
Moves in to a temporary first before copying to the final destination for some reason.

vmovdqa and vmovdqu don't seem to have the same problem.