Improve collation test generation

[echeran]

• Rename collation_short to collation (for test_type) - Test type "collation_short" should be renamed to "collation" throughout the project #186
• For UCA conformance test data, include collation strength IDENTICAL per tuple (according to https://unicode.org/Public/UCA/latest/CollationTest.html)
• For ICU collation test data (collationtest.txt), should create code to parse file in a top-to-bottom stateful-application-of-settings
  • A less-than relation can have an expected strength (<2). This is currently ignored. It can turn into an additional expectation in the test tuple. In this case, the left string should compare secondary less-than the right string. Compare at strength minus 1 (in this case, primary) where they should be equal. At secondary strength, left < right. This would override the general collator strength, with an additional marker to compare equal at the next-lower strength.
  • For tuples from collation conformance test files, ICU needs the normalization=on setting. (Check the ICU test code.) (For tuples from collationtest.txt, this should only be turned on when the test data says so.)
• Fix double escaping problems
  • example executor test case input string: {"label":"0001704","s1":"\\u007f!","s2":"\u0080!","line":339,"ignorePunctuation":true} - note the \\u007f

[sven-oly]

PR #188 addressed double escaping for "s1" and "s2" with big improvements in results in most platforms, e.g. NodeJS 74 from 2,244 to 377 test failures, ICU4C 74 from 6,757 to 4,895 failures.

[sven-oly]

I think this can be marked as "completed".