Generate complete Perl Unicode aliases and wildcard candidates

[fglock]

Problem

PerlOnJava imports current Perl Unicode data, but several generated tables expose only the subset required for direct property lookup. Callers that need value enumeration, especially \p{Property=:\A...\z:} wildcards, can fall back to ICU aliases. ICU may be older than the imported Perl UCD and does not necessarily contain Perl's complete short-alias surface.

The recent Block failures demonstrate the gap: Bengali_Sup, Misc_Arrows_Ext, and Music_Sup are current Perl PropValueAliases.txt aliases, but wildcard matching did not consistently enumerate them.

Proposed work

• Make each generated enumerated-property table expose canonical values and every Perl value alias from the same source data.
• Generate wildcard candidate lists from those aliases, and route wildcard resolution through them ahead of ICU.
• Add generator validation that every accepted alias reaches a generated value set and that aliases are unambiguous.
• Add focused generated-data tests for direct aliases, loose aliases, and anchored wildcards.
• Record which Perl-specific behavior remains handwritten: property grammar, warning categories and wording, case-fold policy, and diagnostics.

Acceptance criteria

• No UCD-version-dependent property lookup or wildcard path uses ICU as its authority when an imported Perl table is available.
• Refreshing imported Unicode data updates value aliases and wildcard coverage without hand-maintained lists.
• Generator checks fail clearly on missing, duplicate, or unreachable aliases.
• The Unicode property UAT remains green for generated direct and wildcard aliases.