Avoid traceback formatting for xfail(run=False)

[ace2016]

Closes #9436.

Summary

Tests marked with xfail(run=False) stop before execution by raising pytest's
internal xfail exception. That exception previously used pytrace=True, so
report generation formatted a full traceback even though the xfail report only
displays the [NOTRUN] reason.

Raise the same internal exception with pytrace=False in both the setup path
and the dynamically evaluated call path. This uses pytest's existing
value-only outcome formatting and leaves imperative pytest.xfail(), normal
failures, and the report pipeline unchanged.

A regression test verifies that xfail(run=False) requests value-only
formatting instead of long traceback formatting.

Testing

• python -m pytest -p no:asyncio testing/test_skipping.py testing/test_reports.py -q
  • 116 passed
• Changed-file pre-commit hooks passed, including Ruff, formatting, codespell,
  and mypy.
• git diff --check passed.

Checklist

• Include documentation when adding new features.
  Not applicable because this is an internal performance improvement.
• Include new tests or update existing tests when applicable.
• Allow maintainers to push and squash when merging my commits.
• Add Closes #9436 to the PR description.
• If AI agents were used, they are credited in Co-authored-by commit trailers.
• Create changelog/9436.improvement.rst.
• Add myself to AUTHORS in alphabetical order.