[CI] UR workflow update

[lukaszstolarczuk]

First part of changes in UR workflow, after moving its content here.

Use detect-changes to only enable UR jobs if UR is affected, plus use the new location of ur project (local instead of clonging UR repo).

[steffenlarsen]

LGTM!

[kbenzie]

Failures are know issues:

• SYCL :: Regression/multithread_write_accessor.cpp timed out in SYCL Pre Commit on Linux / run_prebuilt_e2e_tests #16877
  • https://github.com/intel/llvm/actions/runs/13413586819/job/37471544593?pr=17064
  • https://github.com/intel/llvm/actions/runs/13413586819/job/37471545680?pr=17064
• bindless_images/sampling_3D.cpp and bindless_images/sampling_2D.cpp tests failing with UR_RESULT_ERROR_UNSUPPORTED_FEATURE on HIP/AMD #16933
  • https://github.com/intel/llvm/actions/runs/13413586819/job/37471543244?pr=17064

Battlemage runner seems oversubscribed.

[aelovikov-intel]

Battlemage runner seems oversubscribed.

No, just offline... Second time in two days. I only know how to power cycle the system (done), we'll have to wait for @sarnex to return from his vacation to have a better solution.

However, AMD runner is offline as well:

@kbenzie , can you ping the right folks at CP to look at it?

Same for

since yesterday as well, but at least there is another runner with UR_CUDA label so hasn't become a blocker yet.

[kbenzie]

I've pinged about the cp-amd-runner.
As for the UR_SM_NVIDIA_41_01 runner, I believe that's actually intentional because there were issues with that machine.

[kbenzie]

AMD runner is working, it will however show as offline for a short time between picking up jobs due to how it's been isolated.