reviewer: skip patches meant for xfstests

[OllieinCanada]

Fixes #103

Summary

Skip patches that are clearly intended for xfstests before Sashiko attempts
Linux baseline selection, patch application, or AI review.

Root cause

New xfstests cases commonly add files below tests/<suite>/<case>. Because new
files can apply cleanly to the Linux checkout, Sashiko can mistake the patch
for a kernel change and spend model tokens reviewing it in the wrong
repository.

Changes

• recognize diffs whose changed files are exclusively in the top-level
  tests/<suite>/<path> layout used by xfstests
• mark those patches as skipped before baseline resolution or review work
• complete an all-xfstests patchset without creating an AI review record
• keep Linux selftests under tools/testing/, normal kernel paths, empty
  diffs, and mixed kernel/xfstests diffs in the existing review path

Validation

• focused positive and negative path-classification tests pass
• pipeline test confirms the patch becomes Skipped, the patchset becomes
  Reviewed, and no review record is created
• maintainer-requested path-check simplification passes its focused test
• GitHub Actions DCO, full lint, and Rust unit-test checks pass
• cargo fmt --all -- --check
• git diff --check

[derekbarbosa]

Hi, sorry for the delay!

Thank you for the fix. This looks good. The review comments I posted are minor fixups.

However, this has me thinking if we should try and come up with a solution that would allow certain filepaths to be ignored per-subsystem.

        // Linux selftests live under tools/testing. Top-level tests/<suite>/<case>
        // paths identify xfstests patches, which must not be reviewed as kernel code.

but I am unsure if xfstests' usecase is unique or not. don't consider this a blocker for now :)

[OllieinCanada]

Thanks for the review. I’ve pushed the minor path-check simplification and resolved that thread. I agree that configurable per-subsystem ignored paths may be useful beyond xfstests; since it is not a blocker, I’ve kept this PR narrowly scoped to #103 rather than expanding it here.

[OllieinCanada]

Refreshed this branch against current main and resolved the reviewer test-module overlap while preserving both the xfstests coverage and upstream interaction-ID coverage. The PR is conflict-free and DCO, lint, and unit tests are all green.