Make sure that we vectorize `linalg.pack` + `linalg.unpack` (on top of vectorizing `linalg.mmt4d`) in our [e2e tests](https://github.com/llvm/llvm-project/tree/0d989b2aefe8d8f66981f1d39f56ce0214b5de86/mlir/test/Integration/Dialect/Linalg/CPU) for `linalg.mmt4d`. Specifically, here: * https://github.com/llvm/llvm-project/pull/157815