Fix CPU-GPU linking when ac2g start events are missing.

[devalshahamd]

This pull request improves the robustness of the _find_corresponding_output_event method in trace_to_tree.py by adding a fallback mechanism for runtimes that emit only partial ac2g flow events. The method now attempts a direct correlation-id match when the usual start/end event pairing is incomplete.

Enhanced output event matching logic:

• Added a fallback to match output events by unambiguous correlation-id if only the finish half of the ac2g flow is present, improving compatibility with runtimes like hipDrvLaunchKernelEx.
• Refactored the logic to only require both ac2g start and end events for the primary matching path, and otherwise use the fallback correlation-id matching.

Pull Request Template

Note to AMDers:
This is a public repository. Please do not upload any confidential or customer data. Make sure all such data has been anonymized or removed before making this PR. If you need to attach any private files or links, please insert a Internal OneDrive Link or a Jira Ticket Link instead.