v0.2.2

Other

• Collation for_locale: data-driven from official CLDR rules
• Collation tailoring: unbounded weight allocation (pair-encoded primaries)
• transliteration component delivered (mark done)
• profiling addressed via benchmarks + size tracking
• add Armenian + Georgian romanizations
• Document collation tailoring capacity (gap-insertion limit)
• add alphabetic index (index_labels / index_bucket)
• Format spellout_ordinal regression test (rustfmt)
• Fix RBNF stack overflow / runaway on adversarial input
• add primary-strength string search (find/contains)
• add ordinal spell-out (spell_ordinal)