Fix DateTime test failures: overload warnings and custom warning categories

[fglock]

Summary

This PR fixes DateTime CPAN module test failures by addressing issues with overload warnings, custom warning categories, and lexical warning scope propagation. It also fixes /i flag handling for Unicode properties in extended character classes.

Changes

Overload Warnings

• CompareOperators.java: Add checkSpaceshipResult() to emit "uninitialized value" warning when overloaded <=> returns undef in derived comparison operators (>, <, >=, <=, ==, !=)
• CompareOperators.java: Improve callerWhere() to skip internal Test::* frames for correct warning location reporting

Custom Warning Categories

• warnings/register.pm: Implement proper warnings::register with import() method
• WarningFlags.java: Add registerCategory() for runtime custom warning category registration
• Warnings.java: Add register_categories() method
• ScopedSymbolTable.java: Add registerCustomWarningCategory() for bit allocation

Lexical Warning Scope Propagation

Implement ${^WARNING_SCOPE} mechanism so no warnings 'Category' in user code propagates to warnif() calls in library code (e.g., DateTime.pm).

• WarningFlags.java: Add scope ID tracking infrastructure (registerScopeWarnings(), isWarningDisabledInScope())
• Warnings.java: Register scopes in noWarnings(), check scope in warnIf()
• GlobalContext.java: Initialize ${^WARNING_SCOPE} global variable
• CompilerFlagNode.java: Add warningScopeId field for runtime scope tracking
• StatementParser.java: Pass scope ID to CompilerFlagNode after no warnings
• EmitCompilerFlag.java: Emit local ${^WARNING_SCOPE} = scopeId bytecode
• FindDeclarationVisitor.java: Detect scope nodes for proper local cleanup

Regex /i Flag Fix for Unicode Properties

Fixed the /i flag handling for Unicode properties in extended character classes ((?[...])):

• RegexPreprocessorHelper.java: Wrap Unicode property translations (\p{...}, \P{...}) in (?-i:...) to protect from /i flag
• ExtendedCharClass.java:
  • Thread-local flag to track case-insensitivity through nested processing
  • Extended char class output wrapped in (?-i:...)
  • All literals manually expanded for case-insensitivity
  • Proper handling for interpolated extended char classes with /i flags
  • \N{KELVIN SIGN} and special characters properly case-folded
  • Handle ^ at start of char class as negation, not range start
  • Add LONG S (U+017F) to special case fold mappings
  • Handle \x{...} and \N{...} escapes with case expansion
  • Add special Unicode case folds for ranges like [a-z] with /i
• RegexPreprocessor.java: Skip extended char classes in expandMultiCharFolds(), add \N{...} case-folding

Other Regex and Tied Variable Fixes

• RegexPreprocessor.java: Skip over \p{...}, \P{...}, \N{...}, \x{...}, \o{...} constructs during case-fold expansion
• StringOperators.java: Fix stringConcatWarnUninitialized() to resolve tied scalars once upfront

Reverted Changes (caused regressions)

• TieScalar: Removed set() override that did STORE+FETCH unconditionally
• RuntimeScalar: Reverted FETCH-after-STORE changes for tied scalars

Design Document

See dev/design/warnings-scope.md for the full design and implementation details.

Test Results

All tests match or exceed baselines:

Test	Baseline	Now	Change
op/eval.t	152	153	+1
op/tie.t	45	52	+7
op/bop.t	499	499	-
re/subst.t	184	184	-
re/regex_sets_compat.t	2084	2084	-
re/regex_sets.t	76	81	+5
op/gmagic.t	31	31	-
t/29overload.t	24	27	+3
t/46warnings.t	3	6	+3

Test plan

• Build passes (make)
• All unit tests pass
• No regressions in any of the reported tests
• t/29overload.t passes completely (27/27)
• t/46warnings.t passes completely (6/6)
• /i flag correctly handles Unicode properties
• Interpolated extended char classes preserve /i flag
• Extended char class ranges with /i include special Unicode folds

Generated with Devin