fix: prioritize utf-8 over system locale to prevent crashes on Windows

[CalumRakk]

The Problem

Currently, gitingest crashes (or skips files with an error message) on Windows when processing valid UTF-8 files that contain specific characters (like smart quotes ” or certain math symbols) located past the first 1024 bytes.

Why this happens:

1. The system currently prioritizes locale.getpreferredencoding() (often cp1252 on Windows) over utf-8.
2. cp1252 is "greedy": if the first 1024 bytes of a file are standard ASCII, cp1252 claims the file is valid.
3. However, when reading the full file, if Python encounters a byte that is undefined in cp1252 (e.g., byte 0x9D from a UTF-8 smart quote), it raises a UnicodeDecodeError.

The Solution

I modified src/gitingest/utils/file_utils.py to prioritize utf-8 in the encoding detection list.

Since UTF-8 is stricter than legacy encodings, checking it first ensures that:

1. Valid UTF-8 files are correctly identified immediately.
2. We avoid the "false positive" detection of cp1252.
3. If a file is actually legacy encoding, UTF-8 check will fail quickly, and it will gracefully fallback to the system locale.

Benefits

• CRITICAL: Prevents crashes/read errors on Windows with files containing smart quotes, em dashes, etc.
• Feature: Enables full Emoji support on Windows (prevents "mojibake" like 🚀).
• Consistency: Ensures behavior is identical across Linux, macOS, and Windows.

How to Reproduce

I created a standalone reproduction script. You can run this on Linux/Mac to simulate the Windows environment and witness the crash:

import sys
from pathlib import Path
from unittest.mock import patch

# Add src to path to import internal modules
sys.path.insert(0, str(Path(__file__).parent / "src"))

from gitingest.schemas.filesystem import FileSystemNode, FileSystemNodeType

def reproduction():
    print("--- Simulating Windows (CP1252) Environment ---")

    # Mock Windows environment with CP1252 encoding
    with patch("locale.getpreferredencoding", return_value="cp1252"), \
         patch("platform.system", return_value="Windows"):

        # Create content:
        # 1. Fill initial buffer (1024 bytes) to bypass early detection.
        # 2. Add Emojis and a Smart Quote (”).
        # Note: The smart quote (”) contains byte 0x9D, which is undefined in CP1252 and causes the crash.
        content = (
            "A" * 1050
            + "\n\n"
            + "# Modern Features & Emojis 🚀\n"
            + "This project uses modern features:\n"
            + "- ✨ Magic\n"
            + "- 🔥 Speed\n"
            + "And smart quotes: ” (THIS TRIGGERS THE CRASH)"
        )

        filename = "reproduce_crash.md"
        temp_file = Path(filename)
        temp_file.write_text(content, encoding="utf-8")

        try:
            print(f"Attempting to read '{filename}'...")

            node = FileSystemNode(
                name=filename,
                type=FileSystemNodeType.FILE,
                path_str=str(temp_file),
                path=temp_file,
            )

            result = node.content

            if "Error reading file" in result:
                print("\n[❌ CRASH] Windows failed to read Emojis/Special characters.")
                print(f"Error details:\n{result}")
            else:
                print("\n[✅ SUCCESS] File read correctly!")
                print("Last lines read:")
                print(result[-150:])

        finally:
            if temp_file.exists():
                temp_file.unlink()

if __name__ == "__main__":
    reproduction()

Verification

• Added a regression test tests/test_windows_encoding.py that mocks the Windows environment.
• Verified that existing tests pass.
• Verified code style (pre-commit hooks passed).