Update intel/llvm benchmarks used in CI

[PatKamin]

Description

Update intel/llvm benchmarks framework used to the latest one.
Adjust benchmarking CI to the latest changes regarding data archiving.

This one corresponds with changes up to PR intel/llvm#19128 which introduces a split of benchmark data: current and archived.

Other notable changes:

• updated Compute Runtime to 25.22.33944.8
• chart annotations displaying changes in drivers/benchmarks versions
• added Finalize Graph benchmark (Compute Benchmarks)
• fixed LD_LIBRARY_PATH for Compute Runtime
• skip unpacking of oneAPI if already installed
• more user-friendly Compute Benchmarks charts naming
• improvements in data loading of the dashboard

Checklist

• Code compiles without errors locally
• All tests pass locally
• CI workflows execute properly

• CI workflows, not executed per PR (e.g. Nightly), execute properly

[lukaszstolarczuk]

@PatKamin, why a draft?

And can you please link the PR in the benchmarks (intel/llvm) repo, so I'll know what is this related to.

[PatKamin]

@PatKamin, why a draft?

And can you please link the PR in the benchmarks (intel/llvm) repo, so I'll know what is this related to.

@lukaszstolarczuk, I've updated the PR description