Fix out-of-tree cmake build of torch-mlir-dialects

[oowekyala]

This follows up on #623 for out-of-tree builds of torch-mlir, which added building torch-mir-dialects as a subdirectory.

We found that we needed some more cmake adaptions to make this work. We're initially opening this PR as a draft to gather opinions as to whether this approach suits at least the short-term torch-mlir-dialects goals.

Our goal is to support both in-tree and out-of-tree builds of torch-mlir with minimum hassle, for instance by using the same variable names in both setups. Specifically in externals/llvm-external-projects/torch-mlir-dialects/CMakeLists.txt :

• We use MLIR_FOUND to detect that it is being build as a subdirectory and the llvm+mlir cmake infrastructure is already set up (via find_package in the parent build) as opposed to an in-tree build.
• For in-tree, the setting of variables and loading of llvm+mlir cmake infrastructure is now conditionally performed.
• For in-tree, the names of cmake variables being defined for are adjusted to match those llvm-project makes available through find_package(MLIR REQUIRED CONFIG), under the assumption that those are the more "standardized" names.

Further optional items we could contribute if you're interested:

• build instructions for an out-of-tree setup (eg in README.md)
• a CI flow to test an out-of-tree setup, so that all code paths of the cmake scripts are tested (even though we expect most llvm developers to use in-tree builds)

[silvasean]

Thanks for working on this. Out-of-tree build instructions and CI would be very much appreciated. I'll let @cathyzhyi review the PR itself since she is most familiar with torch-mlir-dialects.

[cathyzhyi]

Thanks for the change! I am not very familiar with llvm cmakes but the change looks reasonable to the best of my knowledge.

Further optional items we could contribute if you're interested:
build instructions for an out-of-tree setup (eg in README.md)
a CI flow to test an out-of-tree setup, so that all code paths of the cmake scripts are tested (even though we expect most llvm developers to use in-tree builds)

Those are great items to add. Thanks!

[oowekyala]

@silvasean @cathyzhyi Thanks for the reviews! I'll put together a followup PR for those additional items.