[llvm] [VectorCombine] Expand `vector_insert` into shufflevector for earlier cost optimizations (#145512) (PR #146479)

[Alexey Bataev via llvm-commits]

alexey-bataev wrote:

> SLP - but I can't remember what the reasoning was why it had to always create vector_insert intrinsics for fixed types - @alexey-bataev can you remember please?

It was just easier to test it, reduces the size of the tests, so we use vector_insert, where possible

https://github.com/llvm/llvm-project/pull/146479