src/sage/tests/books: tweak a "Computational Math..." doctest

One of the tests for the "Computational Math..." book is plotting
lines on a vector field by calling scipy.integrate.odeint repeatedly.
This test runs,

  y = integrate.odeint(f, k, t)
  p += line(zip(t, flatten(y)))

in an attempt to flatten y, which is returned as a numpy array of
1-dimensional numpy arrays. However, flatten() isn't smart enough to
handle that: it returns a list of 1-dimensional numpy arrays rather
than a list of floating point numbers. And numpy-1.25.x now warns
about casting 1-dimensional arrays to float deep in our plotting
infrastructure.

Instead, we now call y.tolist() before passing it to flatten(), and
this does the right thing.

Author: orlitzky

src/sage/tests/books/computational-mathematics-with-sagemath/graphique_doctest.py

@@ -134,11 +134,11 @@
   sage: t = srange(0, 5, 0.1); p = Graphics()
   sage: for k in srange(0, 10, 0.15):
   ....:       y = integrate.odeint(f, k, t)
-  ....:       p += line(zip(t, flatten(y)))
+  ....:       p += line(zip(t, flatten(y.tolist())))
   sage: t = srange(0, -5, -0.1); q = Graphics()
   sage: for k in srange(0, 10, 0.15):
   ....:       y = integrate.odeint(f, k, t)
-  ....:       q += line(zip(t, flatten(y)))
+  ....:       q += line(zip(t, flatten(y.tolist())))
   sage: y = var('y')
   sage: v = plot_vector_field((1, -cos(x*y)), (x,-5,5), (y,-2,11))
   sage: g = p + q + v; g.show()