fix(pytest): handle xfailed summary

[TechWizard9999]

Summary

• fix pytest summary parsing so xfailed is tracked separately instead of being counted as failed
• recognize quiet-mode summary lines like 1 xfailed in 0.01s, which rtk pytest emits because it runs pytest with -q
• return compact success summaries for xfailed-only and mixed passed + xfailed runs
• add regression coverage in src/pytest_cmd.rs and note the fix in CHANGELOG.md

Related Issue

Closes #672

Root Cause / Reproduction

I reproduced this with a minimal pytest suite containing only an @pytest.mark.xfail test.

Raw pytest output:

x                                                                        [100%]
1 xfailed in 0.01s

Before this change, the parser either:

• matched xfailed as failed because it used substring checks, or
• missed the summary line entirely in quiet mode because it only looked for === ... === summaries

That led to incorrect summaries like 1 failed or No tests collected.

Changes Made

• parse summary tokens by exact normalized outcome name instead of substring matching
• add explicit xfailed counting to the pytest summary parser
• detect both classic === ... === summaries and quiet-mode summaries emitted by pytest -q
• keep failure output unchanged while improving success-path summaries

Files Changed

• src/pytest_cmd.rs - fix summary-line detection, parse xfailed, add regression tests
• CHANGELOG.md - add unreleased bug-fix note for xfailed tests reported as '1 failed' in summary output #672

Testing

• cargo fmt --all --check
• rtk cargo build
• rtk cargo clippy --all-targets (reports 2 pre-existing warnings in untouched files: src/gh_cmd.rs, src/git.rs)
• rtk cargo test
• rtk cargo test pytest_cmd
• Manual smoke test with a temporary virtualenv and real pytest xfail case using this branch's built binary

Manual Verification

Using /Users/stack/Documents/rtk/target/debug/rtk pytest against a temp suite containing a single xfail test now prints:

✓ Pytest: 1 xfailed

instead of misreporting the run.

Checklist

• Code follows project style guidelines
• Self-review completed
• Documentation updated
• No breaking changes introduced

cc @pszymkowiak for review