feat(benchmark): add benchmarking framework to establish an FFI migration baseline

[WutangNailao]

Background

To support the upcoming FFI migration in image, we need a repeatable and measurable performance baseline first, so we can evaluate before/after changes with a consistent methodology.

What’s included

This PR introduces benchmark infrastructure:

• Adds a benchmark runner, case registry, API collector, report generator, and image comparison utility.
• Adds multi-resolution samples and standardized output (benchmark/results.json).
• Aligns benchmark cases with exported API symbols to enable coverage-oriented benchmarking.

Usage Examples

# 1) Run benchmarks (generates benchmark/api_manifest.json and benchmark/results.json)
dart run benchmark/runner.dart

# 2) Generate Markdown report (default output: benchmark/benchmark_report.md)
dart run benchmark/report.dart

# 3) Generate report with explicit input/output paths
dart run benchmark/report.dart benchmark/results.json benchmark/benchmark_report.md

# 4) Compare two PNG outputs (optional tolerance, default: 0)
dart run benchmark/compare.dart a.png b.png
dart run benchmark/compare.dart a.png b.png 2

Why this matters

• Establishes a stable performance baseline for future FFI work.
• Enables apples-to-apples before/after performance comparisons.
• Reduces regression debugging cost during migration.

Follow-ups

• Reuse the same benchmark suite in the FFI branch to produce directly comparable results.
• Gradually integrate benchmark runs into CI for continuous performance tracking.