Introduce Rust port of Mini Coding Agent (mini-coding-agent-rs)

[parssky]

Summary

This PR introduces mini-coding-agent-rs , a 1:1 Rust port of the existing Python mini-coding-agent . It brings the exact same robust feature set—including advanced context-reduction, tool sandboxing, and LLM hallucination recovery—while leveraging Rust's performance, safety, and single-binary portability!

Key Implementations & Parity Features:

• Workspace Context ( workspace.rs ): Dynamically captures cwd , branch state, git status, and local docs via process execution to feed context to the LLM immediately.
• Token-Efficient Memory ( session.rs ): Replicates the Python SessionStore JSON architecture, utilizing aggressive clip() strategies and rolling chat history to prevent the LLM context window from overflowing.
• Robust Parsing Engine ( agent.rs ): Implements dual JSON and regex-based XML fallback parsing. If smaller Ollama models generate malformed payloads, the agent intercepts the error and loops back to ask the model to retry.
• Sandboxed Tool Execution ( tools.rs ): Ports all tools ( list_files , read_file , search , write_file , patch_file , run_shell ) with the exact same [y/N] ApprovalPolicy loop to prevent destructive actions from running unchecked.
• CLI & UX ( main.rs ): Fully replicates the /help , /memory , /session , and /reset slash commands alongside the signature ASCII welcome art using clap for identical flag configuration ( --model , --cwd , --approval , etc.).

How to Test

1. Navigate to the mini-coding-agent-rs directory.
2. Run cargo build --release
3. Install globally via cargo install --path .
4. Run mini-coding-agent-rs --model in any project directory and test a /memory or file read prompt!