I'd like to add a small snippet of high-performance non-Python code to an otherwise pure Python package (maintained by someone else). This snippet only contains a few loops over a larger numpy array with only elementary math operations involved (meaning I don't care whether this part is written in C or Fortran or whatever else). Compared to Numpy this snippet does not allocate hidden temporary memory and speeds things up by 10x - 30x.
I'd like to keep distribution of the package as simple as possible and so I'd like to avoid any additional (heavy) dependencies like Numba or relying on Fortran or C compilers. The last drop of performance is not important.
Question: What is the simplest way to include high-performance code snippets into an otherwise pure Python package?