Fix the three findings the 0.22.0 self-check got wrong

[tauanbinato]

The 0.22.0 release self-check baselined four considers as wrong. Three of them came from gaps in JevGate's own policy; this fixes those and measures each fix on the corpus, with every changed review and consider labeled by hand (a debatable one counts as not right).

• Function simplification: splitting a function of 20 lines or fewer is at most a note. Of the considers this lowers, 12 of 39 were right on tuned projects and 5 of 50 on blind Bend 2 projects. Function-simplification considers: 68% to 73% right.
• Shared logic: copies of up to twelve lines between test cases in different files are notes. Of the considers this lowers, 30 of 87 were right on tuned projects and 2 of 24 held out. Shared-logic considers: 53% to 57% right (tuned) and 43% to 53% (held out).
• File organization: a split the recheck raised from an undecided first answer is asked its file's kind, and a kind that serves one feature clears it. 4 of the 18 findings it cleared were right.

Overall, considers went from 57% to 59% right on tuned projects, 47% to 51% held out and 24% to 28% on the blind Bend 2 set. No review changed except one wrong file-organization review, now gone. Rule versions are bumped, and the numbers are in the doc comments and CHANGELOG.

Also:

• unit_outcome is split into a dispatch and helpers (test_pair_outcome, workflows_outcome, staleness_outcome, settings_module_outcome, in_examples). The self-check flagged it once its file changed, and three rules' logic sat inline.
• The laws' "particular inputs" test is now one function shared by the outcome and the wording.

The fourth, test_map::visit, was a split consider no corpus measure separated from right ones (function length, the located block, a whole-body block). It is restructured instead: class_context says whether a class's methods are tests, case what a node is to the walk (a test case, a function that is not one, or something to look inside), and visit only walks. Behavior is unchanged (the test-map tests pass), and the baseline is now empty.

Tests (478), clippy and cargo +1.90.0 check --locked pass. The self-check reports no review or consider.