Add Copilot custom instructions for AceForge context awareness

[Copilot]

Adds .github/copilot-instructions.md to guide Copilot with AceForge-specific context: local-first AI music workstation, macOS/Apple Silicon constraints, PyTorch MPS optimization, and Flask/React architecture patterns.

Contents

• Project context: ACE-Step integration, Demucs, XTTS, basic-pitch; local-first with no cloud dependencies
• Tech stack: Flask blueprints, React 19/TypeScript, Vite, PyTorch with MPS, PyInstaller bundling
• Code conventions:
  • Python: cdmf_* module naming, typed signatures, logging module (not print), Blueprint pattern
  • TypeScript: Functional components, discriminated unions, centralized API layer
• Architecture patterns: Blueprint per feature, lazy model loading, singleton pattern for ML models, MPS device selection
• ML specifics: TORCH_JIT=0 for frozen apps, MPS memory management, 24kHz sample rates, LoRA training
• Platform constraints: Apple Silicon unified memory, ad-hoc code signing, sys._MEIPASS for bundled resources
• Common pitfalls: Don't load models at startup, chunk long audio, handle missing MPS gracefully
• Examples: Blueprint creation with proper logging, React panel patterns

Example pattern

import logging
from flask import Blueprint, jsonify

logger = logging.getLogger(__name__)

def create_feature_blueprint() -> Blueprint:
    bp = Blueprint("feature_name", __name__)
    
    @bp.route("/api/feature", methods=["POST"])
    def handle_feature():
        try:
            data = request.get_json()
            if not data:
                return jsonify({"error": "No data provided"}), 400
            result = process_feature(data)
            return jsonify({"status": "success", "result": result}), 200
        except Exception as e:
            logger.error(f"Feature error: {e}", exc_info=True)
            return jsonify({"error": str(e)}), 500
    
    return bp

404 lines covering security (input validation, sandboxing), performance (MPS optimization, streaming), and testing (pytest, GitHub Actions on macOS runners).

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.