Fix [:C:] getting interpreted as the entire code space

[eggrobin]

@josh-hadley found that after #1057, everything gets U+-escaped in the comments of https://github.com/unicode-org/unicodetools/blob/main/unicodetools/data/security/dev/data/review.txt.

The rule governing what gets escaped is

unicodetools/unicodetools/src/main/java/org/unicode/text/UCD/GenerateConfusables.java

Lines 177 to 178 in 83971c4

	private static final String EXCAPE_FUNNY_RULE =
	":: [[:C:]-[:cn:][:Z:][:whitespace:][:Default_Ignorable_Code_Point:]] hex/unicode ; ";

and the underlying symbol table is the one from TUP:

unicodetools/unicodetools/src/main/java/org/unicode/text/UCD/GenerateConfusables.java

Lines 130 to 137 in 83971c4

	static final UnicodeProperty.Factory ups =
	ToolUnicodePropertySource.make(version); // ICUPropertyFactory.make();
	static {
	// USE the tool unicode set instead of ICU, which may not be using the latest version.
	UnicodeSet.setDefaultXSymbolTable(ups.getXSymbolTable());
	UnicodeTransform.setFactory(TOOL_FACTORY);
	}

ICU UnicodeSet does not give us the possibility of distinguishing \p{X} from \p{X=} in applyPropertyAlias, thus [:C:] is [:C=:].

Recall:

• from UAX44, that ISO_Comment is deprecated and stabilized;
• from Issue 639 Check diffs between TUP and IUP #640 (comment), that TUP has the empty string, rather than null, for its ISO_Comment values;
• from PropertyAliases.txt, that ISO_Comment has the alias ISC;
• from PropertyValueAliases.txt, that the General_Category value Other has the alias C.

Thus [:C=:] is the set of code points with the empty string as their ISO_Comment value (the entire code space according to TUP), rather than the set of code points whose General_Category is a value in the Other grouping (which is [:C:]).

This change fixes the immediate issue, but as noted in the comment, the symbol table used here is dangerous (it caused #1072), and will need to go away as part of #1073 and #1074. I also filed ICU-23087 on the ICU side.

[josh-hadley]

I temporarily patched my branch where I'm working on Confusables with this updated code and it fixes the issue of characters in comments getting replaced with "U+<code>" escapes. So: LGTM, 👍, etc.