Add Dry Run Mode to Culprit Finder

[danielibarrola]

This pull request introduces a Dry Run mode to the Culprit Finder tool. This feature allows users to simulate the bisection process and see exactly which GitHub API calls would be executed without actually triggering workflows, creating branches, or deleting them.

Key Changes

1. Command-Line Interface

• Added a new --dry-run flag to the CLI.
• When enabled, the tool logs simulated actions (e.g., triggering workflows, branch management) instead of performing them.
• Final output (reporting the culprit) is suppressed in dry-run mode to reflect that no real bisection was concluded.

2. DryRunGithubClient Implementation

• Created a DryRunGithubClient class in github_client.py that wraps the standard GithubClient.
• Write Operations: Methods like trigger_workflow, create_branch, and delete_branch are intercepted and replaced with logging statements.
• Read Operations: Methods like compare_commits and get_workflow continue to use the real client to fetch metadata required for the simulation.
• Mocking: Uses unittest.mock to generate simulated WorkflowRun and WorkflowJob objects, allowing the bisection logic to proceed through its state transitions.

3. Core Logic Integration

• Updated culprit_finder.py to accept either the standard or the dry-run client via type hinting.
• In cli.py, the application now conditionally instantiates the DryRunGithubClient based on the user's input.

4. Testing & Documentation

• Added a comprehensive test case in test_culprit_finder.py to ensure that dry-run mode correctly avoids calling write-heavy methods on the real GitHub client.
• Updated README.md to document the new --dry-run flag.

How to use

Run the tool with the --dry-run flag to preview the bisection steps:

culprit-finder <failed-run-url> --dry-run