[CI] Switch to community code formatting check.

[bader]

LLVM project added code format checker - https://github.com/intel/llvm/blob/sycl/.github/workflows/pr-code-format.yml.
I suggest we remove lint check and GitHub Action we developed ourselves. We better contribute/extend community version than maintain our own version.

• pre-commit lint check -

  llvm/.github/workflows/sycl_linux_precommit.yml

  Lines 33 to 60 in 4ca27f8

  	lint:
  	runs-on: [Linux, build]
  	if: ${{ always() && !contains(github.event.pull_request.labels.*.name, 'disable-lint') }}
  	container:
  	image: ghcr.io/intel/llvm/sycl_ubuntu2204_nightly:no-drivers
  	options: -u 1001:1001
  	steps:
  	- uses: actions/checkout@v3
  	with:
  	sparse-checkout: |
  	devops/actions/cached_checkout
  	devops/actions/clang-format
  	devops/actions/cleanup
  	- name: Register cleanup after job is finished
  	uses: ./devops/actions/cleanup
  	- name: 'PR commits + 2'
  	run: echo "PR_FETCH_DEPTH=$(( ${{ github.event.pull_request.commits }} + 2 ))" >> "${GITHUB_ENV}"
  	- uses: ./devops/actions/cached_checkout
  	with:
  	path: src
  	fetch-depth: ${{ env.PR_FETCH_DEPTH }}
  	ref: ${{ github.event.pull_request.head.sha }}
  	merge_ref: ''
  	cache_path: "/__w/repo_cache/"
  	- name: Run clang-format
  	uses: ./devops/actions/clang-format
  	with:
  	path: src

• GitHub Action - https://github.com/intel/llvm/tree/sycl/devops/actions/clang-format

@intel/dpcpp-devops-reviewers.

[aelovikov-intel]

It's running on ubuntu-latest (unless we customize just that piece of it), meaning we might have to wait a lot for its results. I also don't see how we can keep preventing regular CI from starting when the formatting task fails.

[AlexeySachkov]

It's running on ubuntu-latest (unless we customize just that piece of it)

That seems like a reasonable enough downstream customization

I also don't see how we can keep preventing regular CI from starting when the formatting task fails.

AFAIK, there was recently a change made to cancel in-progress CI in case when new commits are pushed to a PR - perhaps that is enough so we can proceed without blocking pre-commit on bad formatting?