fix(python): correct type hint for to_tensor_fn parameter

[AndreaBozzo]

Fixes #3129

Summary

The type hint for to_tensor_fn in LanceDataset.__init__ incorrectly declared a single-argument callable, but the function is called with additional keyword arguments.

Changes

• Updated type hint from Callable[[pa.RecordBatch], ...] to Callable[..., ...] to accurately reflect that the function accepts arbitrary arguments
• Enhanced docstring to document the expected signature with optional hf_converter and use_blob_api keyword arguments

Testing

• Verified Python syntax compilation passes
• No functional changes, only type hint correction

[wjones127]

Can we fix the type hint by making it correct and specific?

[wjones127]

It seems wrong to just make this more generic. How about this?

Callable[[pa.RecordBatch, ...], Union[dict[str, torch.Tensor], torch.Tensor]]

Alternatively, can be fancier and do:

from typing import Protocol

class ToTensorFn(Protocol):
    def __call__(self, batch: pa.RecordBatch, hf_converter, use_blob_api: bool) -> Union[dict[str, torch.Tensor], torch.Tensor]: ...

...
to_tensor_fn: ToTensorFn

[AndreaBozzo]

Thanks for pointing this out, i agree about the hint being too generic.
I modified it with the ' less fancy' approach + ran lint in the last commit