direct numpy conversion and nullspace as matrix

[andibau13]

Hi Alex,

Thanks a lot for creating this tool, it's been very useful for a project related to transversal logic gates! I (or rather Copilot) made two small modifications for my purposes. Wondering if some of this also might be useful to others so I'm sending this pull request - feel free to ignore of course.

(1) I added direct conversion from numpy matrices to bitmatrices. I'm creating the matrices making heavy use of numpy's powerful indexing/slicing methods which BitMatrix doesn't implement, and only performing RREF using your code. Converting very large numpy matrices to BitMatrix via pure-python lists is extremely slow.

(2) I've added a wrapper that returns the nullspace as a matrix, not as a list. More natural for my purposes and it allows me to directly convert to numpy with change (1).

(3) This is not a change, but a small bug I noticed: The kernel of a 0 x n matrix is all n-dimensional space, not empty. Right now, the code returns the empty list. I didn't fix this in the repo but patched it in my python code. I'm wondering if the code would still run without the first equality in the check
'''if self.rows() == 0 || self.cols() == 0 {
return Vec::new();
}
'''
which would fix it.

Best,
Andi