Replace custom `nan_to_num` with Numba's implementation

[Saransh-cpp]

Vector has a custom Numba implementation of numpy.nan_to_num -

https://github.com/scikit-hep/vector/blob/6ba5a60ac90e5afa047f53681257b93ffa158cb5/src/vector/backends/_numba_object.py#L54-L55

which should ideally by in Numba.

The feature request was raised by Jim in https://github.com/numba/numba/issues/6857 and the function was partially implemented in https://github.com/numba/numba/pull/8623. I have opened a new issue for a complete implementation - https://github.com/numba/numba/issues/10089 - and we should be ready to remove our custom implementation once this issue is closed.

[Saransh-cpp]

Blocked by #641