Fix TimeInterpolatedPotential evaluating to zero after pickling

CHANGES.rst

@@ -27,6 +27,9 @@ New Features
 Bug fixes
 ---------
 
+- Fixed ``TimeInterpolatedPotential`` returning zero energy, gradient, and density
+  after being pickled or deep-copied.
+
 - Fixed ``find_actions_o2gf()`` raising an error whenever any toy action or angle
   was NaN. NaN samples are now dropped with a warning, and it raises only if more
   than half of the samples are NaN.

src/gala/potential/potential/builtin/cytimeinterp.pyx

@@ -283,7 +283,7 @@ cdef class TimeInterpolatedWrapper(CPotentialWrapper):
             self.interp_state.wrapped_potential = <void*>self.wrapped_potential.cpotential
 
             self.init(
-                [0.0],  # value of G doesn't matter for this wrapper
+                [G],
                 np.zeros(n_dim, dtype=np.float64),  # placeholder origin
                 np.eye(n_dim, dtype=np.float64),    # placeholder rotation
                 n_dim=n_dim

tests/regression/test_potential_timeinterpolated_539.py

@@ -23,4 +23,10 @@ def test_timeinterpolated_pickle(tmpdir):
         pickle.dump(pot, f)
 
     with open(tmpdir.join("time_interp_pot.pkl"), "rb") as f:
-        pot = pickle.load(f)
+        new_pot = pickle.load(f)
+
+    q = [8.0, 1.0, 0.5] * u.kpc
+    for t in [0.0, 3.3, 10.0] * u.Gyr:
+        assert u.allclose(new_pot.energy(q, t=t), pot.energy(q, t=t))
+        assert u.allclose(new_pot.gradient(q, t=t), pot.gradient(q, t=t))
+        assert u.allclose(new_pot.density(q, t=t), pot.density(q, t=t))