Fix Serbian noun synthesis rules for all 4 declension classes (Vrsta I-IV)

[nciric]

Summary

Implements complete C++ rule-based noun inflection synthesis for Serbian (sr) covering all four primary declension classes (Vrste I – IV) in standard Serbian grammar:

1. Class I (I Vrsta — Genitive Sg -а): Masculine consonant nouns (ученик, прозор, уранак) and unexpanded neuter -о/-е nouns (село, поље, уточиште).
2. Class II (II Vrsta — Genitive Sg -а with stem extension): Neuter -е nouns expanding stem with -н- (име -> имена, време -> времена) or -т- (дете -> детету, дугме -> дугмета, пиле -> пилета).
3. Class III (III Vrsta — Genitive Sg -е): Nouns ending in -а (жена, кућица, Никола, пратња, двојка).
4. Class IV (IV Vrsta — Genitive Sg -и): Feminine consonant nouns ($i$-declension) including primary Instrumental Singular iotation (љубављу, памећу, младошћу, кошћу, ноћу).

Key Refactorings & Optimizations

• Zero-Allocation Case Index Lookup: Replaced runtime std::map heap allocation per call in applySuffix with a constexpr size_t getCaseIndex(std::u16string_view targetCase) helper function (addressing George's feedback in PR Implementing group 3 noun rules for Serbian. #173).
• Disambiguation of -ште Neuters: Excluded Class I -ште neuters (уточиште, положиште) from Class II -ет- stem extension.

Note on Suffix Guessing PR #197

ℹ️ Note: PR #197 (Auto-generated Serbian exemplar suffix mapping table) will be reverted/closed once this C++ synthesis PR lands, as rule-based synthesis now natively covers all four Serbian noun declension classes.

Testing

• Verified with make check -j8 against ICU 78 (263 test cases, 290,506 assertions passed 100%).