[0.2.0] - 2024-06-11

Added - CU Benchmarking Framework

• Systematic CU analysis framework with instruction and transaction paradigms
• Dual benchmarking modes: Pure instruction measurement vs complete transaction workflows
• Statistical analysis engine with percentile-based estimates (min, conservative, balanced, safe, very_high, unsafe_max)
• Rich execution context discovery through simulation (SVM state, program details, CPI analysis)
• Professional tooling integration with env_logger and clean JSON output
• Comprehensive unit tests (324 lines) covering edge cases and percentile calculations

Framework Features

• InstructionBenchmark trait for pure CU measurement without framework overhead
• TransactionBenchmark trait for multi-program workflow analysis
• Two-phase measurement: simulation for context + execution for statistics
• Address book system for human-readable program names
• Type-safe domain modeling with StatType enum

Examples & Documentation

• Working benchmarks: SOL transfer (150 CU), SPL token transfer, token setup workflow (28K-38K CU)
• Comprehensive guide: BENCHMARKING.md with living documentation approach
• Enhanced README: Repositioned as testing + benchmarking platform

Technical Improvements

• Fixed percentile calculation bugs that showed incorrect variance
• Removed automatic ComputeBudgetInstruction for measurement transparency
• SVM state accumulation for realistic vs isolated measurements
• Professional logging (quiet by default, optional progress via RUST_LOG)

Breaking Changes

• None - all existing testing functionality preserved
• New benchmarking features require --features cu_bench opt-in