Implement SQL parser foundation: core types, arena, tokenizer, classifier

[renecannao]

Summary

Build the foundational parser pipeline for a new hand-written recursive descent SQL parser to replace the existing Flex/Bison POC. This covers:

• Core types (StringRef, Dialect enum, StmtType, NodeType)
• Arena allocator (block-chained, zero-copy, sub-microsecond reset)
• Dialect-templated tokenizer (MySQL + PostgreSQL)
• Statement classifier (switch dispatch on first token)
• Tier 2 extractors (table name extraction for DML/DDL, transaction classification, USE database)

Design Spec

docs/superpowers/specs/2026-03-24-sql-parser-design.md

Implementation Plan

docs/superpowers/plans/2026-03-24-sql-parser-foundation.md

Tasks

• Build system setup (Makefile.new, Google Test)
• Core types — StringRef, enums
• Arena allocator
• AstNode (32-byte) and ParseResult
• Token types and keyword tables (MySQL + PostgreSQL)
• Tokenizer (dialect-templated, header-only)
• Classifier and Tier 2 extractors
• Integration smoke test

[renecannao]

Implemented in PR #4. Foundation complete: arena, tokenizer, classifier, Tier 2 extractors, 57 tests.