fix determinant warning

[tylerflex]

I was getting a lot of divide by zero warnings in the np.linalg.det calls from the mode solver.

Digging into the issue, I found it was because taking a determinant of a complex-valued identity matrix triggers this warning, which was the case for the jacobians in the mode solver that were being passed to np.linalg.det.

import numpy as np
arr = np.eye(3) + 0j
np.linalg.det(arr)

/Users/twhughes/.pyenv/versions/3.10.9/lib/python3.10/site-packages/numpy/linalg/linalg.py:2139: RuntimeWarning: divide by zero encountered in det
  r = _umath_linalg.det(a, signature=signature)
/Users/twhughes/.pyenv/versions/3.10.9/lib/python3.10/site-packages/numpy/linalg/linalg.py:2139: RuntimeWarning: invalid value encountered in det

To fix this, I casted the jacobians tonp.real.

@momchil-flex @weiliangjin2021 @dbochkov-flexcompute is this safe you think? or worth worrying about? I tried filtering the warning but don't know if we want to do that more generally in case we should be worrying about this.

[dbochkov-flexcompute]

hmm, interesting, it doesn't complain on my machine for some reason.

but aren't these jacobians supposed to be always real for our purposes? it seems like they are initialized to be complex because they are copies of complex mu_tensor, but maybe we can just do jac_e = np.real(np.copy(mu_tensor))

[tylerflex]

If this is the case, then what you suggest would be ideal. I updated the commit to just cast the jacobians to real when they are copied over. Thanks!