Request for Python version of connected_components and Performance Comparison Inquiry

Hi @ronghanghu

In the previous issues, many people mentioned the installation problems with _C, such as #53 #59 #22 , and I have encountered similar issues:
```
UserWarning: cannot import name '_C' from 'sam2'

Skipping the post-processing step due to the error above.
You can still use SAM 2 and it's OK to ignore the error above,
although some post-processing functionality may be limited
(which doesn't affect the results in most cases; see
https://github.com/facebookresearch/segment-anything-2/blob/main/INSTALL.md).
```
Although it can run ignoring this warning, I still want to understand the impact of this post-processing on the results.

By observing the files in csrc, it was found that there is a function similar to the `union-find algorithm` called `connected_components`.
Could you provide a pure `Python` version of this function? This would be more installation-friendly, especially for people who are not convenient to update the CUDA driver : ).

Or is there a significant difference in time consumption between the CUDA-based implementation and the Python-based implementation? How much exactly is it? In which scenarios would there be a larger difference?

Looking forward to your reply, thank you very much.