[flang] [mlir] [llvm] [openmp] [lldb] [mlir][Vector] Add patterns for efficient i4 -> i8 conversion emulation (PR #79494)

[Diego Caballero via llvm-commits]

dcaballe wrote:

Thanks for the info! I think making the interleave op at Vector level available to fixed vectors would also make sense. There is a point in knowing that a shuffle is actually implementing an interleave pattern.

I guess we should also be fine with this LLVM limitations for now:
```
While this intrinsic supports all vector types the recommended way to express this operation for fixed-width vectors is still to use a shufflevector, as that may allow for more optimization opportunities.
```

Again, if looks like we are building a small ad-hoc backend in here. Ultimately we may want this to be properly supported in LLVM.

https://github.com/llvm/llvm-project/pull/79494