Add Dry Run Mode to Culprit Finder

[danielibarrola]

This PR introduces a dry run mode to the culprit_finder tool. This mode allows users to simulate the bisection process and visualize the steps it would take (branch creation, workflow triggering, and branch deletion) without making any actual state-changing API calls to GitHub.

Key Changes

Core Logic (culprit_finder.py)

• Added a dry_run parameter to the CulpritFinder constructor.
• Updated _test_commit to skip triggering GitHub workflows when dry_run is enabled, returning True by default to allow the bisection loop to proceed.
• Modified run_bisection to:
  • Log "DRY RUN" messages instead of calling create_branch and delete_branch.
  • Skip the standard "Commit is good/bad" status logging to avoid confusion with the mock results.

CLI Interface (cli.py & README.md)

• Added a new --dry-run flag to the CLI.
• When --dry-run is used, the script will exit after the simulation without attempting to print a final culprit (since no real testing occurred).
• Updated the documentation to include the new flag and its usage.

Testing (test_culprit_finder.py & test_cli.py)

• Added comprehensive unit tests for the dry run mode:
  • test_run_bisection_dry_run: Verifies that no state-changing API calls (create_branch, delete_branch, trigger_workflow) are made.
  • test_test_commit_dry_run: Ensures _test_commit behaves correctly in dry run mode.
• Updated existing CLI tests to account for the new constructor signature.

How to use

Run the culprit finder with the --dry-run flag:

python -m culprit_finder.cli --repo owner/repo --start <good_sha> --end <bad_sha> --workflow ci.yml --dry-run