Sort CMake configuration options and update actions/checkout

[keryell]

This allows easier comparisons between the various CI configurations.
Use latest GitHub actions/checkout@v4
This fixes some deprecation warnings about old NodeJS.
Also fix some trailing space issues by side effect.

[keryell]

@makslevental @fifield It is unclear to me what is happening here.
I this my fault or just some external planet misalignment?

[makslevental]

https://github.com/Xilinx/mlir-aie/actions/runs/8398472772/job/23003372175?pr=1151#step:5:2301

Cloning into '/home/runner/work/mlir-aie/mlir-aie'...
fatal: Remote branch sort-cmake-options not found in upstream origin
Error: Process completed with exit code 128.

I have said this many many many times: there is no point in people that are part of the org (acdc) sending PRs from forks.

Beyond that I would appreciate it if you actually reverted the changes to mlirDistro.yml and mlirAIEDistro.yml (i.e., the two actions that triggered this fail) - they build the distro releases that downstream users depend on (i.e., IREE) and changing the ymls triggers rebuilds.

Finally, the orders of the CMake options weren't arbitrary - they were grouped by (CMake, LLVM, MLIR, AIE). I'm not opinionated about it but to me it was more important to distinguish between generic options (CMake), upstream (MLIR/LLVM) and our options.

[keryell]

I have said this many many many times: there is no point in people that are part of the org (acdc) sending PRs from forks.

Sorry, Alzheimer is hitting me at my age. :-(
But otherwise, repository hygiene and preparing for community open-source inclusion come to my mind.

Beyond that I would appreciate it if you actually reverted the changes to mlirDistro.yml and mlirAIEDistro.yml (i.e., the two actions that triggered this fail) - they build the distro releases that downstream users depend on (i.e., IREE) and changing the ymls triggers rebuilds.

Done. Interestingly this was just fixing some NodeJS deprecation.

Finally, the orders of the CMake options weren't arbitrary - they were grouped by (CMake, LLVM, MLIR, AIE). I'm not opinionated about it but to me it was more important to distinguish between generic options (CMake), upstream (MLIR/LLVM) and our options.

I can understand but it is unclear this was fully the case and lexicographic sort is not that bad since the hierarchical prefixes enforce some locality even if it not perfect for some concepts crossing prefixes.
What does the team think about it?