Automate running benchmarks in PRs that are more likely to impact performance

[celinval]

Proposed change:

• Automatically run performance benchmark workflow if the PR code has changed and the PR has the "Z-benchCI" label.
• Automatically tag PRs with "Z-benchCI" if they touch files are likely to impact Kani's performance. For example:
  • kani-compiler/**
  • rust-toolchain.toml
  • kani-dependencies
  • kani-driver/src/call-*

Motivation: After #2604, running benchmark workflow for a PR will be triggered manually. In order to reduce the risks of forgetting to do so, we should automate running the workflow in riskier scenarios.

My suggestion is to use the Z-benchCI as the source of truth for whether the workflow runs or not. This will allow us to override the automation. For example, if the changes only touches comments of a compiler file.

[celinval]

See https://github.com/marketplace/actions/labeler for that

[celinval]

It looks like the auto label logic doesn't work to trigger other github actions. See: https://github.com/orgs/community/discussions/33804

[celinval]

Just to clarify what's going on. The github documentation has a good explanation on why the auto label workflow does not trigger a 'labeled' action:

When you use the repository's GITHUB_TOKEN to perform tasks, events triggered by the GITHUB_TOKEN, [redacted] , will not create a new workflow run.

To fix this issue, we have the following options:

• Create a new token and use that instead.
• Explicitly trigger the workflow when the label is added.