fix: Asynchronous Pipeline Support and explicit use of utf-8 encoding when reading module files

[lizhenisu]

The Issue

Fix: Force UTF-8 encoding when loading pipeline modules to resolve UnicodeDecodeError on Windows

On Windows, the default encoding for open() is often system-dependent (e.g., GBK or CP1252). If a user writes a custom pipe function containing UTF-8 characters (like Chinese prompts), the loader fails to read the file, raising a UnicodeDecodeError.

Error Log:

UnicodeDecodeError: 'gbk' codec can't decode byte 0x97 in position ...: illegal multibyte sequence

Reproduction Steps

1. On a Windows machine.
2. Create a custom pipeline file with the following content (containing Chinese characters):

# ... inside the pipe function
ollama_messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "把图片内容转成文字"}, # <--- This triggers the error
            image_part
        ]
    }
]
# ...

[WhatIsData]

@tjbck Would you be able to handle the pr for giving support to allow the pipe within Pipeline to be async?