Add comprehensive benchmarking suite

[muk2]

Summary

• Adds criterion-based benchmark suite covering the full SQL processing pipeline
• Introduces src/lib.rs to expose library API for benchmark access
• Benchmarks span 13 SQL query categories from simple SELECTs to complex CTEs

Benchmark Groups

Group	What it measures
parsing	SQL text → AST conversion across 13 query types
compilation	AST → SQL text generation
optimization	AST transformation passes (redundant nesting removal)
round_trip	Full parse → optimize → compile pipeline
multi_statement	Parsing 2 and 5 statements in a single input
explain	EXPLAIN plan text parsing (small and large plans)

Query Categories Benchmarked

• Simple SELECT, SELECT with WHERE, JOINs, multi-JOIN (4 tables)
• Aggregation with GROUP BY/HAVING
• CTEs (WITH clause, 2 CTEs)
• Window functions (ROW_NUMBER, AVG, SUM with frames)
• Subqueries (IN, EXISTS, scalar)
• UNION ALL (3-way)
• INSERT, UPDATE, DELETE
• CASE expressions

How to Run

cargo bench                    # Run all benchmarks
cargo bench -- parsing         # Run only parsing benchmarks
cargo bench -- round_trip      # Run only round-trip benchmarks

Test plan

• All 225 existing tests pass
• Benchmarks compile (cargo bench --no-run)
• cargo build succeeds
• Library crate properly exposes needed modules

Closes #46

🤖 Generated with Claude Code