backend: support non-ASCII characters in path to llmodel libs on Windows

[cebtenzzre]

There were two problems preventing GPT4All from locating the implementation libraries on Windows if the path contained non-ASCII characters:

• We were using LoadLibraryExA (which normally only supports ASCII) and not LoadLibraryExW
• std::filesystem doesn't understand UTF-8 on Windows and must be used with wide strings instead

While I was touching this code I refactored Dlhandle into a proper .cpp/.h split, which greatly improves readability IMO.

Related to #2111 (that issue was about the model path, not the implementation lib path)

Testing

I created a venv called zz😊 (for tab completion's sake) using pure Python (since using Unicode at the Windows command line is not trivial):

$ python
>>> import sys
>>> sys.argv[1:] = ['zz\U0001F60A']
>>> import venv.__main__

Then I activated it (the glob allows this to work even when the console doesn't fully support Unicode):

$ . .\zz??\Scripts\Activate.ps1

Then install the Python binding in that venv. With the current release, this fails:

$ python
>>> from gpt4all import GPT4All
>>> x = GPT4All('orca-mini-3b-gguf2-q4_0.gguf') 
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\msys64\home\jared\gpt4all\zz😊\Lib\site-packages\gpt4all\gpt4all.py", line 206, in __init__
    self.model = LLModel(self.config["path"], n_ctx, ngl)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\msys64\home\jared\gpt4all\zz😊\Lib\site-packages\gpt4all\_pyllmodel.py", line 218, in __init__
    raise RuntimeError(f"Unable to instantiate model: {'null' if s is None else s.decode()}")
RuntimeError: Unable to instantiate model: Could not find any implementations for build variant: default

But with this PR (don't install it editable/-e or the resolved path won't be in the venv), it works:

$ python
>>> from gpt4all import GPT4All
>>> x = GPT4All('orca-mini-3b-gguf2-q4_0.gguf')
>>>

I didn't test with the UI since the installer doesn't let you select a path containing non-ASCII characters, but if the user moves GPT4All to a non-ASCII path after-the-fact, this should at least give it a chance of working.