Fix coating brownian so it agrees with Hong

The small remaining discrepancy between Hong and gwinc is due to an error in gwinc's implementation of the Yam et al model.

The coefficient in front of the phase derivative dcdp_z is supposed to be nN (as written correctly in Yam et al, eq 1), not nN/2 (as found in gwinc).

In Hong et al, the coefficient is nN/2, but that's because their phases are defined as one-way instead of round-trip (see Yam et al, eq A2).

After correcting the error, gwinc and Hong agree, as expected, for the case of equal bulk and shear losses and no photoelasticity.

Source of the plot

gwinc/nb.py

@@ -252,7 +252,8 @@ class Budget(Noise):
         # initialize all noise objects
         for nc in self.noises:
             name = self.__init_noise(nc, noises)
-            self._budget_noises.add(name)
+            if name:
+                self._budget_noises.add(name)
         # initialize common calibrations and add to all budget noises
         for cal in self.calibrations:
             self.__add_calibration(cal, self._budget_noises)
@@ -270,25 +271,21 @@ class Budget(Noise):
         cal = None
         if isinstance(nc, (list, tuple)):
             noise = nc[0]
-            cals = nc[1]
+            cals = nc[1:]
         else:
             noise = nc
             cals = []
-        if noise_filt and noise not in noise_filt:
-            return
-        name = self.__add_noise(noise)
-        for cal in cals:
-            self.__add_calibration(cal, [noise])
-        return name
-
-    def __add_noise(self, noise):
         noise_obj = noise(
             *self.args,
             **self.kwargs
         )
         name = noise_obj.name
+        if noise_filt and name not in noise_filt:
+            return
         logging.debug("init {}".format(noise_obj))
         self._noise_objs[name] = noise_obj
+        for cal in cals:
+            self.__add_calibration(cal, [name])
         return name
 
     def __add_calibration(self, cal, noises):