Notebook hangs when repeatedly calling Python extension module function

This code runs fine in command-line ipython, but breaks Jupyter notebook (same virtual env):

```python
import numpy as np

import cdyn

c = 0.251 + .0001j
maxiter = 10000
res = 1000
extent = (.32, .62, -.27, .03)

x = np.linspace(extent[0], extent[1], num=res+1)
y = np.linspace(extent[2], extent[3], num=res+1)

for i in range(res+1):
print(i) # added for debugging
for j in range(res+1):
cdyn.test_fulljulia(x[i], y[j], c, maxiter)
```

I call a C function from a custom extension module 1,000,000 times (once in each iteration of a double loop of 1000 iterations each). For the purposes of debugging, I'm ignoring the return value of `cdyn.test_fulljulia`, but the original script that reads those values successfully plots a Julia set image.

When I run the code above in command-line ipython, it always works fine and takes around 30 seconds. However, when I run the exact same code in Jupyter notebook, the kernel freezes, and the only way to get rid of the `[*]` status indicator is to shutdown the kernel (pressing `ii` to send an interrupt signal does not work), close the notebook tab in Firefox, and open it again.

I don't think the C function is the culprit. It is actually very simple, just math, no memory allocation at all (except for a `return PyLong_FromLong(i);` at the end). It's in this gist https://gist.github.com/fofoni/cc9fda75298f2394a17ef42f5e9b7935 if anyone wants to take a look.

When I add the `print(i)` line as in the code above, I can see that multiple iterations of the loop are successfully run before it freezes. The number of times the kernel calls `cdyn.test_fulljulia` looks random to me (each time I close/reopen the notebook and try again, it's a different number), but usually 10,000 to 100,000 times (that is, the last `print(i)` I see is between 10 and 100). Indeed, if I tune down that `res`olution parameter, then the Jupyter notebook that plots the Julia set works just fine.