Ignore np.log(np.nan)

This happens when initialising. Since it results in NaN anyway, we should be able to detect other errors downstream.

Author: hyanwong

tsdate/base.py

@@ -125,7 +125,7 @@ def force_probability_space(self, probability_space):
             if self.probability_space == LOG:
                 pass
             elif self.probability_space == LIN:
-                with np.errstate(divide="ignore"):
+                with np.errstate(divide="ignore", invalid="ignore"):
                     self.grid_data = np.log(self.grid_data)
                     self.fixed_data = np.log(self.fixed_data)
                 self.probability_space = LOG