[llvm] [LAA][NFC] Add a test for #191867 (PR #213722)

[Ramkumar Ramachandra via llvm-commits]

https://github.com/artagnon commented:

Not sure if this test makes sense: there's a stride1 variant, followed by a stride2 variant, followed again by a stride1 variant; why would the second stride1 variant be influenced by the stride2 variant? Was it happening before your fix?

https://github.com/llvm/llvm-project/pull/213722