bench(rust): add criterion.rs benchmarks for evaluation, operators, and state management

[aepfli]

Summary

• Add criterion.rs as dev dependency with HTML report generation
• Create 3 benchmark suites covering the core Rust evaluation engine
• All benchmarks compile and run via cargo bench

Benchmark Suites

benches/evaluation.rs — Flag Evaluation (8 benchmarks)

Benchmark	What it measures
evaluate_flag_simple	Boolean flag, no targeting
evaluate_flag_targeting_match	Targeting rule that matches
evaluate_flag_targeting_no_match	Targeting rule, default path
evaluate_flag_complex_targeting	Nested and/or with multiple conditions
evaluate_flag_disabled	Disabled flag (early return)
evaluate_flag_not_found	Missing flag key (error path)
evaluate_logic_simple	Direct JSON Logic {==: [1,1]}
evaluate_logic_complex	Complex rule with custom operators

benches/operators.rs — Custom Operators (8 benchmarks)

Benchmark	What it measures
fractional/buckets/2	A/B test bucketing
fractional/buckets/4	4-way bucketing
fractional/buckets/8	8-way bucketing
semver_equals	Version equality
semver_range/caret	Caret range matching
semver_range/tilde	Tilde range matching
semver_range/gte_prerelease	Greater-than-or-equal with prerelease
starts_with / ends_with	String prefix/suffix matching

benches/state.rs — State Management (5 benchmarks)

Benchmark	What it measures
update_state/flags/5	Small config load
update_state/flags/50	Medium config load
update_state/flags/200	Large config load
update_state_no_change	Re-apply identical config
update_state_incremental	Change 1 flag in 100-flag config

How to run

# Run all benchmarks
cargo bench

# Run specific suite
cargo bench --bench evaluation
cargo bench --bench operators
cargo bench --bench state

# Quick run (fewer iterations)
cargo bench -- --quick

# HTML reports generated in target/criterion/

Test plan

• All 3 benchmark suites compile
• cargo bench runs successfully
• cargo fmt and cargo clippy pass
• Review HTML reports in target/criterion/

Closes #63

🤖 Generated with Claude Code