fix: patch OpenAI client in distil module to support dispatch mode

[divyanshupatel17]

Describe your changes

• Fixed a bug in instructor.distil where the OpenAI client was not being patched with instructor, causing dispatch mode to fail because response_model was not recognized.
• Added comprehensive unit tests for instructor.distil in tests/test_distil.py covering key functionality including get_signature_from_fn, format_function, and the @distil decorator in both usage modes.
• Refactored imports in distil.py to use top-level relative imports to avoid potential circular dependencies.

Issue ticket number and link

(None)

Checklist before requesting a review

• I have performed a self-review of my code
• If it is a core feature, I have added thorough tests.
• If it is a core feature, I have added documentation.

[jxnl]

I wonder if the right thing to do here is to actually allow the distillation tool to support any client that is an instructor client. This was implemented when opening; that was the only option, but we can go beyond this now.

[divyanshupatel17]

Thanks for the feedback @jxnl !

Looking at "examples/distilations/three_digit_mul_dispatch.py" ,
I see it already passes an Instructor client. Should I update to:

1. Accept Union[OpenAI, Instructor] and only patch raw clients?
2. Also rename openai_client -> client for clarity?
3. Or keep the current fix minimal (just patching) and leave broader changes for a follow-up?
   Happy to go with whichever approach fits the project best.