Expose coredump check artifacts from guest-side collection

[saisharan0103]

The coredump check was collecting crash reports and core files on the guest during after-test checks, then discovering result log entries by running find over the shared checks directory and filtering each path with runner-side Path.exists(). Because that happened before the second post-check pull copied the newly created guest files back, collected coredump files were dropped from results. The same discovery path could also include files owned by other checks that had already been pulled and publish the internal coredump-latest marker.

This changes _get_crashes to track the exact report and core paths it creates and return them with the non-ignored crash records. _check_coredump now converts only those owned paths to paths relative to invocation.phase.step_workdir, appends failures.yaml, and no longer uses find plus runner-side existence checks for result discovery.

The dump.*.txt command also now invokes coredumpctl info directly instead of through an invalid sh -c wrapper, preserving sudo, options, PID, and redirection so the saved text file contains details for the selected crash.

ruff check tests/test/check/test-coredump.sh tmt/checks/coredump.py reports no new findings on the changed files.
Ran pytest -x locally with no new failures.