Implement native function of `initcap`

[slfan1989]

Summary

Add a native, vectorized initcap(str) that capitalizes the first letter of each word and lowercases the rest, matching Apache Spark semantics.

Motivation

• Align with Spark SQL string functions to improve coverage.
• Use an Arrow-native implementation to boost performance on large columns.

Scope

• New scalar function: initcap(str)
• Input types: STRING (Utf8, LargeUtf8)
• Output type: STRING
• Null semantics: initcap(NULL) → NULL
• Empty string returns empty string

Expected Semantics (Spark-aligned)

• “Word” boundaries are split by non-letter characters.
• For each word: uppercase the first letter, lowercase the remaining letters.
• Non-letter characters (spaces, punctuation, digits, underscores, hyphens, etc.) are preserved as-is and act as delimiters.
• Unicode-friendly: letters include non-ASCII letters (accents, CJK, etc.).
• Locale-independent: use Unicode case mapping; not affected by system locale.