Add JIT mode architecture proposal and performance analysis

[Copilot]

Proposes JIT compilation architecture to address interpreter overhead in hot validation paths.

Analysis

Current bottlenecks (4-step schema):

• 12 function calls per execution (proxy → wrapper → isSuccess → step logic)
• 4 intermediate result object allocations
• Array iteration through runtimeSteps on every execute()
• Try-catch wrapper on every step

JIT potential: 3-6x speedup via code generation eliminating loop unrolling, function boundaries, and intermediate allocations.

Proposal Documents

• jit-mode-support.md: Four implementation strategies (function generation, lazy compilation, AOT API, hybrid fast paths) with 3-phase rollout recommendation
• jit-performance-analysis.md: Execution path analysis, memory characteristics, compilation cost modeling (break-even at ~10 executions)
• jit-mode-discussion-zh-tw.md: Chinese translation for maintainer
• README.md: Proposals directory index and contribution template

Recommended Path

1. Phase 1: Monomorphic optimizations in createPipeExecutor, fast paths for common primitives → 20-30% gain
2. Phase 2: Explicit .compile() API using new Function() → 3-6x gain
3. Phase 3: Auto-compilation after threshold → transparent optimization

Open Questions

• Opt-in vs automatic compilation
• Async step handling strategy (exclude, AsyncFunction, or mixed)
• CSP fallback approach
• Compilation threshold tuning
• Compiled schema chainability trade-offs

Configuration

Updated eslint.config.js to ignore docs/proposals/*.md for code example flexibility.

Original prompt

建立一個 issue 用來和我討論目前的程式碼架構要怎麼樣有效的新增支援 JIT 模式

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.