Adds support for reordering=False in Butina clustering

[Matthew-Neba]

Summary

Adds support for reordering=False in Butina clustering. Fixes #192

Changes

• Added a reordering option to the Python and C++ Butina APIs.
• Implemented fixed-order Butina clustering for reordering=False.
• Added tests and ensured they passed
• Updated the benchmark to optionally run nvMolKit Butina with reordering enabled, disabled, or both.

[scal444]

Thanks for the contribution! This looks great, and we can definitely get it in for v0.6. A few comments, mostly minor.

[Matthew-Neba]

Thanks for all the help! I have fixed the issues you pointed out. I have managed to parallelize that kernel and fixed the other minor issues aswell. I could indeed use the blockReduce idea you mentioned and it works well.

[scal444]

Looks great! Can you install ruff 0.15.8 and clang-format-17, and then run admin/run_clang_format.sh and admin/run_ruff_format.sh to get it to pass the linters?

[scal444]

Or i can do it and push a commit if you're ok with that

[Matthew-Neba]

No worries, I ran the formatters. I also fixed an issue that was leading to the forge build failing. Hopefully all loose ends are tied now.