Initially it was planned as tests for #101372. That bug occurred because many tests were only ran with the current Unicode database, not UCD 3.2.0. So, I made most tests running with both databases, added tests for normalization, added many examples with Unicode >4.0.0 in other tests, and marked Unicode version for existing examples. Since running new tests takes time, I optimized tests, slow tests which was previously skipped without the "cpu" resource, now run with limited range. Now the run without the "cpu" resource takes 3 seconds (was 1 second), and full run -- 30 seconds (was 17 seconds).

Linked PRs

• gh-142681: Improve unicodedata tests #142682
• [3.14] gh-142681: Improve unicodedata tests (GH-142682) #142708
• [3.13] gh-142681: Improve unicodedata tests (GH-142682) (GH-142708) #142710
• gh-142681: Move NormalizationTest-3.2.0.txt to more safe place. #142935
• gh-142681: Preserve Lib/test/data/NormalizationTest-3.2.0.txt in make distclean #142943
• [3.14] gh-142681: Move NormalizationTest-3.2.0.txt to more safe place. (GH-142935) #142944
• [3.13] gh-142681: Move NormalizationTest-3.2.0.txt to more safe place. (GH-142935) #142945